{"schema_version":1,"benchmark":"JevBench by Benchmark Heaven","version":"v1.6.1","published_at":"2026-10-06T00:39:36.000Z","source_sha256":"5cd8c1332226ea9c179883d2285c0d8b8cc593ae38f2645197364a7ddd9c83c1","method":{"capability":{"definition":"Arithmetic mean of Intelligence and Calibration; ranked models within both official caps. Headline on the open-weights board.","reference_key":"jev-1.13.0@v1.5.7","cost_cap_usd_per_1000":0.06459465517241379,"median_latency_cap_s":1.2329566404223442,"factor":2},"composite":{"definition":"Weighted harmonic mean of Intelligence, Calibration, Speed and Cost, multiplied by squared penalties when Intelligence is below its configured floor, or Speed or Cost is below 50. Equal 25% weights do not mean an arithmetic average.","weights":{"calibration":25,"cost":25,"intelligence":25,"speed":25},"intelligence_floor":50,"speed_floor":50,"cost_floor":50,"zero_axis_score":0,"missing_axis_score":null}},"scope":"Published v1.6.1 including addenda a6, a7, a8, a11, a12 through 2026-10-08; cite source_sha256 for the exact result bytes. Later live-board reruns and dated carry remain separate.","rows":[{"key":"sage-1.3.0","name":"Sage 1.3.0 (Levanto Labs)","board":"api","model_pin":"levanto-sage-v1.3 (Sage 1.3.0)","last_measured_on":"2026-10-05","source_url":"https://levanto.ai","listing":"ranked","ranked":true,"capability":78.59554275566781,"capability_eligible":true,"open_capability_rank":null,"composite":74.04992179175468,"intelligence":65.59290944525489,"calibration":91.59817606608073,"speed":93.61638181239104,"cost_axis":58.231203787388296,"usd_per_1000_decisions":0.024677048070413,"price_kind":"estimate","price_basis":"Levanto list tariff, top tier: USD 0.05 per 1M input tokens, USD 10 per 1M output tokens (https://levanto.ai/pricing, read 5 Oct 2026)","p50_raw_s":0.15245595201849937,"p50_adjusted_s":0.15245595201849937,"latency_adjustment":"none (API)"},{"key":"h2o-lightning-4b","name":"H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim)","board":"open","model_pin":"h2oai/h2o-lightning-4b@193ad740925b176a3b70a5a13a7cff2f2fadd01e","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/h2oai/h2o-lightning-4b","listing":"ranked","ranked":true,"capability":74.99619636935047,"capability_eligible":true,"open_capability_rank":5,"composite":72.5232271521435,"intelligence":60.03839906196252,"calibration":89.95399367673842,"speed":92.61084474720658,"cost_axis":60.29882443070202,"usd_per_1000_decisions":0.021055837438423645,"price_kind":"estimate","price_basis":"ESTIMATE (self-hosted open weights, no public tariff): cost per 1,000 decisions measured for H2O-Lightning-4B v1.0 on the v1.5 pool (same base, shim and token path) carried for v1.1, the practice of every Qwen3.5-4B row on this board; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off. Cross-check: v1.1's own v1.6 input tokens on the common cost basis give USD 0.0204 per 1,000 (within 3 %)","p50_raw_s":0.02928461297415197,"p50_adjusted_s":0.20856922594830393,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"mercury-decide","name":"Mercury Decide (Inception; System One decisions API, served free on OpenRouter as inception/mercury-decide:free)","board":"api","model_pin":"provider-reported measurement model slug: inception/mercury-decide:free","last_measured_on":"2026-10-07","source_url":"https://openrouter.ai/inception/mercury-decide:free","listing":"ranked","ranked":true,"capability":73.62521498320774,"capability_eligible":true,"open_capability_rank":null,"composite":72.36824049637728,"intelligence":65.90701345473542,"calibration":81.34341651168008,"speed":85.47752366478576,"cost_axis":62.08313746631354,"usd_per_1000_decisions":0.018360961408259987,"price_kind":"estimate","price_basis":"LABELLED ESTIMATE, not charged - and nothing WAS charged: usage.cost is 0 on every one of the 1,500 calls. Inception publishes NO paid tariff for Mercury Decide: OpenRouter lists it free-only (prompt 0, completion 0) and the non-free slug inception/mercury-decide has no endpoints at all (404 'No endpoints found', read 7 Oct 2026), so there is nothing bookable. Because the system is API-only with undisclosed weights, the open rows' size-class estimate table cannot apply either. The row is therefore priced from THE SAME OPERATOR'S cheapest published per-token tariff on the SAME platform, inception/mercury-2.5 at USD 0.04 per 1M input and USD 0.15 per 1M output, which IS a key in the frozen 25 Sep snapshot - x this row's own measured tokens. Output tokens are 3 per decision on average (4,434 over 1,477 answers) because the probability is read from the model rather than written out, so the output rate is nearly inert and the Cost axis is set by the input rate alone. RELEASE-LANE LINE, recorded and not acted on, and it is RUN 57'S SHAPE EXACTLY - one mapping line, not a new price: the rate inception/mercury-2.5 (0.04, 0.15) is already IN the dated snapshot; what is missing is only a base-model decision line pointing at it, so pricing_v15.reference() raises on every inception/* string. receipts/COST-ALTERNATIVES-R58.json has the two alternatives' arithmetic already done.","p50_raw_s":0.33213604614138603,"p50_adjusted_s":0.33213604614138603,"latency_adjustment":"none (API)"},{"key":"decisio-gemma-4-31b-v080","name":"decisio v0.8.0 on gemma-4-31B-it (frozen, FP8 on load, one prefill per state, self-hosted)","board":"open","model_pin":"aminry/decisio@5b42101a191d222062a795194b0aedb5003bf6ea + google/gemma-4-31B-it@842da379","last_measured_on":"2026-10-06","source_url":"https://github.com/aminry/decisio","listing":"ranked","ranked":true,"capability":79.56974174390604,"capability_eligible":true,"open_capability_rank":2,"composite":71.6926576652985,"intelligence":70.42321091828492,"calibration":88.71627256952715,"speed":91.18580810090454,"cost_axis":51.66530928015878,"usd_per_1000_decisions":0.04084664184157075,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff) and CLEAN: unlike run 54's FP8 sibling, the frozen pricing snapshot resolves this EXACT base - google/gemma-4-31B-it - to USD 0.09 per 1M input and USD 0.34 per 1M output. Read before the pod and asserted in make_registry_r55.py. x this row's own measured tokens, the deck31b method. Nothing is provisional and no release-lane line is open on it. Estimated, not charged","p50_raw_s":0.047096215188503265,"p50_adjusted_s":0.24419243037700653,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jev-1.13.0","name":"Jev 1.13.0 (TypeSafe AI)","board":"reference","model_pin":"jev-1.13.0 (response model id on all 590 successful rows)","last_measured_on":"2026-10-05","source_url":"https://docs.typesafe.ai","listing":"ranked","ranked":true,"capability":77.12647993588574,"capability_eligible":true,"open_capability_rank":null,"composite":71.49118639284514,"intelligence":63.62075784024779,"calibration":90.63220203152369,"speed":91.49989299046987,"cost_axis":54.733842199536014,"usd_per_1000_decisions":0.03227542180094787,"price_kind":"tariff","price_basis":"TypeSafe list tariff USD 0.042 per 1M input tokens, output free; measured on the v1.6.1 common cost basis (all items except those outside the Jev reference input range) from this run's token usage","p50_raw_s":0.23910215869545937,"p50_adjusted_s":0.23910215869545937,"latency_adjustment":"none (API)"},{"key":"quyet-1-0-large","name":"Quyet-1.0-Large (Chinh Nguyen, Gemma-4-31B decoder, option-letter logits)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/chinhnc/Quyet-1.0-Large","listing":"ranked","ranked":true,"capability":81.68880454670384,"capability_eligible":true,"open_capability_rank":1,"composite":71.38706329141472,"intelligence":73.40135940607429,"calibration":89.97624968733336,"speed":86.87440406834797,"cost_axis":50.546097335562465,"usd_per_1000_decisions":0.04451059729064039,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-31b-it list price $0.09/M input, $0.34/M output (the exact base model google/gemma-4-31B-it; listed since Apr 2026), read 2026-10-04 x the package's own measured usage.input_tokens; one forward pass, nothing generated; run in process on 1x H100 80GB (Lium), quyet 1.0.0; estimated, not charged","p50_raw_s":0.11577823059633374,"p50_adjusted_s":0.3815564611926675,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decider-12b","name":"decider-12b v2 (Mapika)","board":"open","model_pin":"Mapika/decider-12b@8ac1efa708b71b86ae33b01d2a8d7a3ddcb48e66","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/Mapika/decider-12b","listing":"ranked","ranked":true,"capability":72.50047627291733,"capability_eligible":true,"open_capability_rank":8,"composite":70.87955164338175,"intelligence":63.01250112161242,"calibration":81.98845142422223,"speed":90.47583056481872,"cost_axis":57.755449421401664,"usd_per_1000_decisions":0.025594796798029562,"price_kind":"estimate","price_basis":"add-requests run 46 (v1.5 basis, unpublished) documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.054131726268678904,"p50_adjusted_s":0.25826345253735783,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"wity-1","name":"wity-1 (Wity, reasoning auto)","board":"api","model_pin":null,"last_measured_on":"2026-10-06","source_url":"https://wity.alphanimble.com/","listing":"ranked","ranked":true,"capability":79.11559492280433,"capability_eligible":false,"open_capability_rank":null,"composite":70.76534010331063,"intelligence":70.28143483604077,"calibration":87.9497550095679,"speed":71.79000307789984,"cost_axis":58.8349720563948,"usd_per_1000_decisions":0.023559582938388626,"price_kind":"tariff","price_basis":"Wity's stated API tariff USD 0.042 per 1M input tokens, output free; measured on the v1.6.1 common cost basis (all items except those outside the Jev reference input range) from this run's token usage","p50_raw_s":1.5746641159057617,"p50_adjusted_s":1.5746641159057617,"latency_adjustment":"none (API)"},{"key":"decider-12b-v1","name":"decider-12b v1, stock Gemma-4-12B-it (Mapika)","board":"open","model_pin":"Mapika/decider-12b@09e9e3387fa42fed740233be2290385d18a6044d","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/Mapika/decider-12b","listing":"ranked","ranked":true,"capability":72.1959364423342,"capability_eligible":true,"open_capability_rank":10,"composite":70.42352588714175,"intelligence":60.62893554403879,"calibration":83.7629373406296,"speed":90.47654290454523,"cost_axis":57.755449421401664,"usd_per_1000_decisions":0.025594796798029562,"price_kind":"estimate","price_basis":"add-requests run 46 (v1.5 basis, unpublished) documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.054043288342654705,"p50_adjusted_s":0.25808657668530943,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"torchcast-decision-12b","name":"torchcast-decision-12b (Torchcast AI, Gemma-4-12B fine-tune, option-letter logprob readout)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/torchcast-ai/torchcast-decision-12b","listing":"ranked","ranked":true,"capability":71.73894611323924,"capability_eligible":true,"open_capability_rank":11,"composite":69.9087143535159,"intelligence":60.54320091058725,"calibration":82.93469131589123,"speed":91.74573681567192,"cost_axis":56.36034169226667,"usd_per_1000_decisions":0.02848756157635468,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; no public tariff, customer asked for the base-model reference): OpenRouter google/gemma-3-12b-it list price $0.05/M input, $0.15/M output (gemma-4-12b-it is not listed on OpenRouter, read 2026-10-04; same reference as the Cygnet and Jev-Omni Gemma-4-12B rows) x the shim's own measured usage.prompt_tokens/completion_tokens; self-hosted on 1x H100 80GB (Lium), vLLM 0.30.0 BF16; estimated, not charged","p50_raw_s":0.037151537369936705,"p50_adjusted_s":0.2243030747398734,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"winnow-12b","name":"Winnow-12B Q8","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/EldanRing/Winnow-12B","listing":"ranked","ranked":true,"capability":71.24586176250037,"capability_eligible":true,"open_capability_rank":13,"composite":68.88542163459404,"intelligence":59.53366229166759,"calibration":82.95806123333315,"speed":86.69753831262085,"cost_axis":56.556023424877196,"usd_per_1000_decisions":0.028062900246305422,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.1168902744539082,"p50_adjusted_s":0.38378054890781643,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"deck31b","name":"deck-31B (krishna765, frozen Gemma-4-31B-it, TorchAO FP8 dynamic)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://github.com/krishna-gogineni-765/deck31b","listing":"ranked","ranked":true,"capability":77.5907429125261,"capability_eligible":true,"open_capability_rank":3,"composite":68.68200118286292,"intelligence":73.02512554551848,"calibration":82.15636027953371,"speed":86.69970536196755,"cost_axis":49.69190284016272,"usd_per_1000_decisions":0.04752658090724441,"price_kind":"estimate","price_basis":"ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-31b-it list price $0.09/M input, $0.34/M output (the exact frozen base model google/gemma-4-31B-it; same reference as Quyet-1.0-Large), read 2026-10-04 x the server's own measured usage.input_tokens; letter-mass readout, nothing generated; TorchAO FP8 dynamic on 1x GPU (Lium), deck31b 1.0.0; estimated, not charged","p50_raw_s":0.12746820924803615,"p50_adjusted_s":0.4049364184960723,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"cygnet","name":"Cygnet (blockbrain, frozen Gemma-4-12B-it)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/blockbrain-ai/cygnet-recipe","listing":"ranked","ranked":true,"capability":70.90341055515565,"capability_eligible":true,"open_capability_rank":14,"composite":68.55012424151383,"intelligence":54.83928122751086,"calibration":86.96753988280044,"speed":91.77283664982562,"cost_axis":56.42912569481113,"usd_per_1000_decisions":0.028337561576354687,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.03681807057000697,"p50_adjusted_s":0.22363614114001393,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decisio-gemma-4-12b-v080","name":"decisio v0.8.0 on gemma-4-12B-it (frozen, one forward pass, self-hosted)","board":"open","model_pin":"aminry/decisio@5b42101a191d222062a795194b0aedb5003bf6ea + google/gemma-4-12B-it@707f0a3b","last_measured_on":"2026-10-06","source_url":"https://github.com/aminry/decisio","listing":"ranked","ranked":true,"capability":70.74146614369144,"capability_eligible":true,"open_capability_rank":15,"composite":68.17346094077045,"intelligence":52.324988528776075,"calibration":89.15794375860679,"speed":86.9860428376769,"cost_axis":59.34583145470655,"usd_per_1000_decisions":0.022653689911983754,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): the exact base google/gemma-4-12B-it is in the frozen pricing module's UNLISTED_BASE_MODELS set, so no exact base-model rate and no floor exist. The reference is the one the rows PUBLISHED on this same base use - cygnet, jev-omni and winnow-12b - OpenRouter google/gemma-3-12b-it, the nearest publicly hosted 12B Gemma sibling, at its FULL catalog pair USD 0.05 per 1M input and USD 0.15 per 1M output. PREREG-R54-AMENDMENT-1 uses the full pair rather than run 53's output=0, so a row that emits tokens pays for them; where a row generates nothing the output term is zero anyway. x this row's own measured input/output tokens, the deck31b method. Estimated, not charged","p50_raw_s":0.10244052350753918,"p50_adjusted_s":0.3548810470150784,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decisio-gemma-4-12b-v090","name":"decisio v0.9.0 on gemma-4-12B-it (frozen, one prefill per question, self-hosted)","board":"open","model_pin":"aminry/decisio@333a0a849de7658969a8f8933b34a8097a8c3489 + google/gemma-4-12B-it@707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7","last_measured_on":"2026-10-07","source_url":"https://github.com/aminry/decisio","listing":"ranked","ranked":true,"capability":69.95438022832886,"capability_eligible":true,"open_capability_rank":18,"composite":67.76072338665118,"intelligence":52.299106200484864,"calibration":87.60965425617285,"speed":85.86778537116794,"cost_axis":59.34583145470655,"usd_per_1000_decisions":0.022653689911983754,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff), CARRIED AND NOT NEW: the exact base google/gemma-4-12B-it is in the frozen pricing module's UNLISTED_BASE_MODELS set, so no exact base-model rate and no floor exist, and (0.05, 0.15) is the reference the rows PUBLISHED on this same base already use - cygnet, jev-omni, winnow-12b and torchcast-decision-12b - OpenRouter google/gemma-3-12b-it, the nearest publicly hosted 12B Gemma sibling, at its FULL catalog pair USD 0.05 per 1M input and USD 0.15 per 1M output. It is also the rate THIS ROW'S OWN v0.8.0 SIBLING carries, so the two versions are priced identically and the comparison between them is clean. PREREG-R54-AMENDMENT-1's full pair rather than output=0, x this row's own measured input/output tokens, the deck31b method. The server reports its own usage.input_tokens/output_tokens on every answer (664,759 in and 1,477 out over the answered rows: one prefill, one option-label readout step per decision), so no token-accounting amendment was needed. Estimated, not charged.","p50_raw_s":0.13179851847235113,"p50_adjusted_s":0.4135970369447023,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jev-omni","name":"Jev-Omni (akhilaaa3, Gemma-4-12B merged)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/akhilaaa3/Jev-Omni","listing":"ranked","ranked":true,"capability":71.2873662785918,"capability_eligible":true,"open_capability_rank":12,"composite":67.74758858028281,"intelligence":55.530225828826175,"calibration":87.04450672835743,"speed":85.43173201875288,"cost_axis":56.05113155939739,"usd_per_1000_decisions":0.02917173645320197,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.15469176834449172,"p50_adjusted_s":0.45938353668898346,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"xor-26b-a4b","name":"Xor 26B-A4B (Juspay, Gemma-4-26B-A4B, bf16)","board":"open","model_pin":"juspay/Xor-26B-A4B@8afd4396492084c7c91846ed3791161d54665d34","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/juspay/Xor-26B-A4B","listing":"ranked","ranked":true,"capability":72.40651985976965,"capability_eligible":true,"open_capability_rank":9,"composite":67.3818936950357,"intelligence":58.673823517178164,"calibration":86.13921620236111,"speed":91.03179229063062,"cost_axis":50.69588844260597,"usd_per_1000_decisions":0.04400179417738659,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-26b-a4b-it in the frozen 25 Sep snapshot, the M2 reference of the published surogate-rune-26b-a4b-v3 row (the frozen module has no explicit listing-decision line for this base: release-lane line), USD 0.09 per 1M input and USD 0.3 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.058566353749483824,"p50_adjusted_s":0.26713270749896767,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"bobcat-flash-1.2","name":"Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted)","board":"open","model_pin":"sanghwa-na/bobcat-flash-1.2@58d15456937ce58625b13a5e3d982846554a3aef","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/sanghwa-na/bobcat-flash-1.2","listing":"ranked","ranked":true,"capability":73.4936722737144,"capability_eligible":true,"open_capability_rank":6,"composite":67.34032787154446,"intelligence":65.81464576505633,"calibration":81.17269878237248,"speed":90.781726629608,"cost_axis":49.68577376542501,"usd_per_1000_decisions":0.04754894380501015,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-26b-a4b-it in the frozen 25 Sep snapshot, USD 0.09 per 1M input and USD 0.3 per 1M output x this row's own measured tokens. Estimated, not charged. THIS IS THE REFERENCE, THE RATE AND THE LABEL THE PUBLISHED PAGE ALREADY CARRIES FOR THIS EXACT BASE on three rows published within the last day - xor-26b-a4b, surogate-rune-26b-a4b-v3-v16 and gev-26b-decide - so pricing this row any other way would make it inconsistent with the field it is placed in. RELEASE-LANE LINE, recorded and not acted on: pricing_v15.reference() RAISES on the author's exact declared spelling google/gemma-4-26B-A4B-it, because that string is in none of the three decision tables, while the lowercase google/gemma-4-26b-a4b-it IS in BASE_REFERENCES and in the dated snapshot at (0.09, 0.30). That is a CAPITALISATION gap - one notch tighter than run 56's -Base/instruct gap on Qwen3.5-35B-A3B - and the three published rows say in their own basis string that the frozen module has no explicit listing-decision line for this base. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: their build_response returns usage.output_tokens == 0 and the readout takes candidate logits at the first answer position of one forward pass, so this row's cost depends on the input rate alone and any other decision is reversible by arithmetic from the raw.","p50_raw_s":0.060955737018957734,"p50_adjusted_s":0.2719114740379155,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"spx-cd-flash","name":"SPX-CD Flash (SurdAI, hosted /v1/systemone, Oct-4 checkpoint)","board":"api","model_pin":"provider-reported measurement model slug: spx-cd-flash","last_measured_on":"2026-10-06","source_url":"https://surd.ai","listing":"ranked","ranked":true,"capability":74.31544056451752,"capability_eligible":true,"open_capability_rank":null,"composite":66.68017155386212,"intelligence":58.43051843707937,"calibration":90.20036269195566,"speed":79.39117060033084,"cost_axis":52.107175358413464,"usd_per_1000_decisions":0.039484576844955994,"price_kind":"estimate","price_basis":"Operator's own public-beta overage tariff, conservative reading USD 0.04 per 1M input tokens, output USD 0 (the provider's /v1/models payload reports output_price \"0\"): three dated primary readings of the operator's own published overage tariff: 24 Sep 2026 provider console (flash 0.025, pro 0.08 per 1M input), 1 Oct 2026 docs page (flash 0.04, pro 0.02), 6 Oct 2026 docs page AND /v1/models payload agreeing (flash 0.02, pro 0.04). Sources conflict, so the per-model MAXIMUM is used, which is the published fastino-gliner-2-5-decide precedent (\"published sources conflict ... conservative higher\"). Today's agreeing pair would be USD 0.02/M, i.e. half this row's cost. No base-model floor applies: the submitter states the base model stays sealed until a later open release, so no base model exists to reference. Priced on usage.input_tokens as the provider reports them. The provider also reports billable_input_tokens = input_tokens / 2 on every single item (297/297 of the public set, exactly 2.0): its default effort 2 renders the prompt twice and its docs bundle the second pass free during public beta (\"default effort 2 is included at the single-pass input price\"). That bundle is a public-beta promotion, and TASK.md forbids a free or promotional tier setting the price, so the two passes actually consumed are charged. The operator's own billed figure is exactly half of this row's cost","p50_raw_s":0.9804960675537586,"p50_adjusted_s":0.9804960675537586,"latency_adjustment":"none (API)"},{"key":"decider-chat-gemma4-31b","name":"decider chat on Gemma-4-31B-it (Mapika, frozen base, inference technique)","board":"open","model_pin":"google/gemma-4-31B-it@842da3794eaa0b77d5f08bae87a17459d91ff475 + Mapika/decider-chat-gemma4-31b@22975cbd1716170264a1d2a7e1105787e08c5d95","last_measured_on":"2026-10-06","source_url":"https://github.com/Mapika/decider","listing":"ranked","ranked":true,"capability":70.58267678405554,"capability_eligible":true,"open_capability_rank":16,"composite":66.07956202755186,"intelligence":58.54290392078144,"calibration":82.62244964732962,"speed":85.78163968405963,"cost_axis":50.78502198104071,"usd_per_1000_decisions":0.04370179417738659,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-31b-it, the reference of the published deck31b and quyet-1-0-large rows, USD 0.09 per 1M input and USD 0.34 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.12556729599600658,"p50_adjusted_s":0.4011345919920132,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"hopper-12b-trained","name":"Hopper 12B trained (gemma-4-12B-it frozen + LoRA r32 unmerged, one forward pass, self-hosted)","board":"open","model_pin":"HopitAI/hopper@4b605f25eb903c2cf4d121040ffb097240b3bc37","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/HopitAI/hopper","listing":"ranked","ranked":true,"capability":68.14610127602074,"capability_eligible":true,"open_capability_rank":19,"composite":65.78419365065008,"intelligence":51.6576814732076,"calibration":84.63452107883388,"speed":84.57301814857726,"cost_axis":56.15749203571245,"usd_per_1000_decisions":0.02893456330399459,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): the exact base google/gemma-4-12B-it is in the frozen pricing module's UNLISTED_BASE_MODELS set, so no exact base-model rate and no floor exist. The reference is the one the rows PUBLISHED on this same base use - cygnet, jev-omni and winnow-12b - OpenRouter google/gemma-3-12b-it, the nearest publicly hosted 12B Gemma sibling, at its FULL catalog pair USD 0.05 per 1M input and USD 0.15 per 1M output. PREREG-R54-AMENDMENT-1 uses the full pair rather than run 53's output=0, so a row that emits tokens pays for them; where a row generates nothing the output term is zero anyway. x this row's own measured input/output tokens, the deck31b method. Estimated, not charged","p50_raw_s":0.1638150625221897,"p50_adjusted_s":0.4776301250443794,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"surogate-rune-26b-a4b-v3-v16","name":"Surogate Rune 26B-A4B v3 (v1.6 pool)","board":"open","model_pin":"surogate/rune-26b-a4b-GGUF@bd4a7cbbed66af1a95dfbac12cabf52e6e677411","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/surogate/rune-26b-a4b-GGUF","listing":"ranked","ranked":true,"capability":73.46807972140657,"capability_eligible":true,"open_capability_rank":7,"composite":65.0490738113794,"intelligence":56.055735926690254,"calibration":90.88042351612289,"speed":86.95393114544444,"cost_axis":49.597326937281316,"usd_per_1000_decisions":0.047872830060934325,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-26b-a4b-it in the frozen 25 Sep snapshot, the M2 reference of the published surogate-rune-26b-a4b-v3 row (the frozen module has no explicit listing-decision line for this base: release-lane line), USD 0.09 per 1M input and USD 0.3 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.1161980559991207,"p50_adjusted_s":0.3823961119982414,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"gev-26b-decide","name":"GEV-26B-Decide (AutoTrust, Gemma-4-26B-A4B + LoRA + head)","board":"open","model_pin":"autotrust/GEV-26B-Decide@7c89590ead085bf77630b4bf68264ea30b6ddc78","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/autotrust/GEV-26B-Decide","listing":"ranked","ranked":true,"capability":70.07613110259072,"capability_eligible":true,"open_capability_rank":17,"composite":64.2201412904201,"intelligence":49.817428608862244,"calibration":90.33483359631919,"speed":89.87310530072351,"cost_axis":51.12030271996105,"usd_per_1000_decisions":0.04259153012863913,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter google/gemma-4-26b-a4b-it in the frozen 25 Sep snapshot, the M2 reference of the published surogate-rune-26b-a4b-v3 row (the frozen module has no explicit listing-decision line for this base: release-lane line), USD 0.09 per 1M input and USD 0.3 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.051788969489280134,"p50_adjusted_s":0.2535779389785603,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"spx-cd-omni","name":"SPX-CD-Omni (SurdAI, google/gemma-4-12B-it + LoRA r32, one forward pass, self-hosted)","board":"open","model_pin":"SurdAI/SPX-CD-Omni@06a786a13f2ce9b8d1b8f541cf030fae007008fb + google/gemma-4-12B-it@707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/SurdAI/SPX-CD-Omni","listing":"ranked","ranked":true,"capability":68.1286624051734,"capability_eligible":true,"open_capability_rank":20,"composite":63.16406085537614,"intelligence":48.779308810848946,"calibration":87.47801599949784,"speed":89.44891259349035,"cost_axis":58.27072766098771,"usd_per_1000_decisions":0.024602301963439405,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): the exact base google/gemma-4-12B-it is in the frozen pricing module's UNLISTED_BASE_MODELS set, so no exact base-model rate and no floor exist. The reference is the one the three rows PUBLISHED on this same base all use - cygnet, jev-omni and winnow-12b - OpenRouter google/gemma-3-12b-it list price USD 0.05 per 1M input, USD 0.00 per 1M output (the nearest publicly hosted 12B Gemma sibling; output is 0 because the runner reads candidate logits in one forward pass and generates nothing), x this row's own measured usage.input_tokens, the deck31b method. Estimated, not charged","p50_raw_s":0.0733477434841916,"p50_adjusted_s":0.2966954869683832,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"diffusion-jev","name":"Diffusion Jev (DiffusionGemma 26B-A4B on patched SGLang, 48-step diffusion readout, self-hosted)","board":"open","model_pin":"google/diffusiongemma-26B-A4B-it@f7f5b7f5fa82ffc52addd066915886d497f5517b","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/google/diffusiongemma-26B-A4B-it","listing":"ranked","ranked":true,"capability":58.08750812795111,"capability_eligible":true,"open_capability_rank":42,"composite":59.330428224994314,"intelligence":54.860121638521335,"calibration":61.31489461738089,"speed":86.79193858661813,"cost_axis":49.559714426383636,"usd_per_1000_decisions":0.04801123222748815,"price_kind":"estimate","price_basis":"ESTIMATE, and the caveat travels with the number rather than sitting in a footnote. The frozen pricing snapshot returns None for google/diffusiongemma-26B-A4B-it (it is unlisted: no exact rate and no floor), and a scan of all 129 registry entries finds NO published row on any DiffusionGemma base at all - the page's openjev-razorback16 is the same model but a different submitter and carries no v1.6 score - so there was no precedent to copy. The reference used is google/gemma-4-26b-a4b-it, USD 0.09 per 1M input and USD 0.30 per 1M output: the identically sized and identically shaped publicly hosted Gemma 4 26B-A4B sibling of the unlisted diffusion checkpoint, a closer sibling than the one run 54 had to use for gemma-4-12B-it. x this row's own measured tokens. THE CAVEAT: a token-priced cost axis charges this row as though it ran one forward pass, when the author's own documentation denoises a 256-position canvas for up to 48 adaptive steps per question. The Cost axis therefore UNDERSTATES this configuration's compute, by construction and not by accident; the measured per-question denoising_steps are kept in the raw so the understatement is quantifiable. Estimated, not charged","p50_raw_s":0.1306727642659098,"p50_adjusted_s":0.4113455285318196,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"rene-1-31b-fp8","name":"René-1 31B FP8 (salfatigroup, Gemma 4 31B full fine-tune, one-pass option readout)","board":"open","model_pin":"salfatigroup/rene-1-31b-fp8@bd634489957f8da63ccce858ee33fd6c2929d286","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/salfatigroup/rene-1-31b-fp8","listing":"ranked","ranked":true,"capability":75.9540833183004,"capability_eligible":true,"open_capability_rank":4,"composite":55.84876743320955,"intelligence":61.72139577013068,"calibration":90.18677086647014,"speed":87.11720932683018,"cost_axis":45.963638345890566,"usd_per_1000_decisions":0.06327207176709546,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference, not charged). Unmodified pricing_v15.reference resolves google/gemma-4-31B-it to google/gemma-4-31b-it at USD0.09/M input and USD0.34/M output in frozen snapshot. Measured output tokens are 0 by construction (decision_contract.py and release.json); one forward-pass option readout. Common v1.6.1 cost input range applied unchanged.","p50_raw_s":0.10970725817605853,"p50_adjusted_s":0.3694145163521171,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"aplomb-1","name":"Aplomb 1 (5.3B decision model on Qwen3.5-4B, trained readout head, self-hosted)","board":"open","model_pin":"empiriolabsai/aplomb-1@f8e2dc81f0505db21a3b7a933ea479658e4ebb8f","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/empiriolabsai/aplomb-1","listing":"ranked","ranked":true,"capability":67.51828279493242,"capability_eligible":true,"open_capability_rank":21,"composite":54.55838408944504,"intelligence":45.81854920739271,"calibration":89.21801638247211,"speed":88.88310989631262,"cost_axis":57.86508463809988,"usd_per_1000_decisions":0.025380324983073795,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights): the exact declared base Qwen/Qwen3.5-4B resolves in the frozen pricing snapshot through DEEPINFRA_BASE_REFERENCES to USD 0.03 per 1M input and USD 0.15 per 1M output, read before the pod and asserted in make_registry_r55.py. The same rate run 54 gave messier-one, the other self-hosted Qwen3.5-4B derivative. DELIBERATELY NOT the vendor's own cheaper public tariff: Aplomb 1 is also sold on EmpirioLabs' hosted API at USD 0.02 per 1M input with output free (GitHub #195), which would LOWER this row's cost - but all 106 self-hosted rows in the registry carry cost_kind `estimate` and not one is priced at a vendor tariff, and creating that precedent for one row is the release lane's call, not this job's. The estimate can therefore only OVERSTATE this row, and it is reversible by arithmetic because the raw keeps the tokens. x this row's own measured tokens. INPUT TOKENS ARE EXACT, NOT ESTIMATED, and they include BOTH passes of the author's own --debias flag on every multi-option choice, because that is what the configuration consumes. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: the author's answer_logits runs the model with logits_to_keep=1, use_cache=False and reads label logits, and nothing is generated anywhere in their code path. Counts carried in the scorer's own cost_estimate field, never written into usage. Estimated, not charged","p50_raw_s":0.09363547945395112,"p50_adjusted_s":0.33727095890790226,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"bespoke-nimble-9b-v3","name":"Bespoke Nimble 9B v3 (Qwen3.5-9B LoRA)","board":"open","model_pin":"bespokelabs/Bespoke-Nimble-9B-v3@8e927b9b4afdbb14479fac10a7364d1a695be208","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/bespokelabs/Bespoke-Nimble-9B-v3","listing":"ranked","ranked":true,"capability":66.79290880490993,"capability_eligible":true,"open_capability_rank":22,"composite":53.47656993239491,"intelligence":56.54011796945148,"calibration":77.04569964036838,"speed":89.19502471208428,"cost_axis":46.094951684283686,"usd_per_1000_decisions":0.06263757616790792,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter qwen/qwen3.5-9b, the reference of the published nimble-9b row (same base), USD 0.1 per 1M input and USD 0.15 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.09388846100773662,"p50_adjusted_s":0.33777692201547327,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"apus-openjev-v1-9b","name":"APUS-OpenJev-v1-9B (merged bf16 checkpoint-3000 on Qwen3.5-9B, their own native runtime, full depth, self-hosted)","board":"open","model_pin":"apus-ailab/APUS-OpenJev-v1-9B@82c9c56cfa9de8d36704ed91948d4726ef111635","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/apus-ailab/APUS-OpenJev-v1-9B","listing":"ranked","ranked":true,"capability":56.907845099085506,"capability_eligible":true,"open_capability_rank":47,"composite":49.17976892452783,"intelligence":47.20514429531085,"calibration":66.61054590286017,"speed":82.18245466956745,"cost_axis":48.738123569705756,"usd_per_1000_decisions":0.05113628977657414,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights): the exact declared base Qwen/Qwen3.5-9B resolves in the frozen pricing snapshot through BASE_REFERENCES to qwen/qwen3.5-9b = USD 0.10 per 1M input and USD 0.15 per 1M output, read before the pod and asserted in make_registry_r56.py. Nothing provisional. x this row's own measured tokens. BECAUSE OUTPUT TOKENS ARE 0 BY CONSTRUCTION this axis rests on the input rate alone. Estimated, not charged.","p50_raw_s":0.277784826583229,"p50_adjusted_s":0.705569653166458,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"plumb-4b","name":"Plumb-4B (crh225, JevK5 v0.2 + LoRA)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/crh225/plumb-4b","listing":"ranked","ranked":true,"capability":65.42184422831923,"capability_eligible":true,"open_capability_rank":23,"composite":48.28224126395837,"intelligence":42.9626698747687,"calibration":87.88101858186977,"speed":93.83356539601934,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.019995882059447467,"p50_adjusted_s":0.18999176411889493,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"spx-cd-pro","name":"SPX-CD Pro (SurdAI, hosted /v1/systemone, Oct-4 checkpoint)","board":"api","model_pin":"provider-reported measurement model slug: spx-cd-pro","last_measured_on":"2026-10-06","source_url":"https://surd.ai","listing":"ranked","ranked":true,"capability":78.63514339971708,"capability_eligible":false,"open_capability_rank":null,"composite":47.66970820706367,"intelligence":64.85693822910903,"calibration":92.41334857032513,"speed":77.96292698098645,"cost_axis":43.076275488494026,"usd_per_1000_decisions":0.07896915368991199,"price_kind":"estimate","price_basis":"Operator's own public-beta overage tariff, conservative reading USD 0.08 per 1M input tokens, output USD 0: three dated primary readings of the operator's own published overage tariff: 24 Sep 2026 provider console (flash 0.025, pro 0.08 per 1M input), 1 Oct 2026 docs page (flash 0.04, pro 0.02), 6 Oct 2026 docs page AND /v1/models payload agreeing (flash 0.02, pro 0.04). Sources conflict, so the per-model MAXIMUM is used, which is the published fastino-gliner-2-5-decide precedent (\"published sources conflict ... conservative higher\"). Today's agreeing pair would be USD 0.04/M, i.e. half this row's cost. No base-model floor applies: the submitter states the base model stays sealed until a later open release. Priced on usage.input_tokens as the provider reports them. The provider also reports billable_input_tokens = input_tokens / 2 on every single item (297/297 of the public set, exactly 2.0): its default effort 2 renders the prompt twice and its docs bundle the second pass free during public beta (\"default effort 2 is included at the single-pass input price\"). That bundle is a public-beta promotion, and TASK.md forbids a free or promotional tier setting the price, so the two passes actually consumed are charged. The operator's own billed figure is exactly half of this row's cost","p50_raw_s":1.1281277425587177,"p50_adjusted_s":1.1281277425587177,"latency_adjustment":"none (API)"},{"key":"messier-one-v0.2","name":"Messier One v0.2 (Qwen3.5-4B fine-tune, one prefill, self-hosted)","board":"open","model_pin":"agentmessier/messier-one@8bbddb84140e8fc732b35713f998ed628fc70306","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/agentmessier/messier-one","listing":"ranked","ranked":true,"capability":58.130562443744644,"capability_eligible":true,"open_capability_rank":41,"composite":45.29059491870636,"intelligence":42.35139759222662,"calibration":73.90972729526267,"speed":92.70909408602279,"cost_axis":64.78240136302922,"usd_per_1000_decisions":0.014925192958700068,"price_kind":"estimated","price_basis":"ESTIMATE (base-model reference; self-hosted open weights). NO JUDGEMENT OF OURS: pricing_v15.reference() resolves Qwen/Qwen3.5-4B straight through to deepinfra:Qwen/Qwen3.5-4B at USD 0.03 per 1M input and USD 0.15 per 1M output in the frozen 25 Sep snapshot (the same rate aplomb-1 carries for the same base) x this row's own measured tokens. Estimated, not charged. THE FRONT END REPORTS NO TOKEN USAGE (usage {}), exactly as v0.1 did in run 54, so the counts come from run 54's pre-registered route unchanged (PREREG-R59-AMENDMENT-1 = R54-AMENDMENT-2): the author's own prompt.Prompter (v0.2 source, unmodified) recomputes each item's prompt token ids, output is fixed at 1 by their own max_tokens=1, and the totals were required to match vLLM's own counters or the row would get no cost: 914,665 input and 1,500 output tokens both ways (vllm prompt_tokens_total 915,349 - 684; generation_tokens_total 1,508 - 8). Counts sit in cost_estimate, never in usage.","p50_raw_s":0.027263494484941475,"p50_adjusted_s":0.20452698896988294,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"spark-s1-4b-v6","name":"spark-s1-4b-v6 (Open Spark Jev, abhishek085)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://github.com/abhishek085/open-spark-jev","listing":"ranked","ranked":true,"capability":54.016043277778124,"capability_eligible":true,"open_capability_rank":53,"composite":44.90217636528639,"intelligence":43.2364603010071,"calibration":64.79562625454915,"speed":86.95618624402076,"cost_axis":60.422063196854964,"usd_per_1000_decisions":0.020857610837438423,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.1286082404985791,"p50_adjusted_s":0.40721648099715824,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"swanone","name":"swanOne (blockbrain, Qwen3.8-Flash-Next NVFP4)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/blockbrain-ai/swanone-recipe","listing":"ranked","ranked":true,"capability":72.36627654122822,"capability_eligible":false,"open_capability_rank":null,"composite":43.78640157551217,"intelligence":56.15104294255194,"calibration":88.58151013990448,"speed":82.64878475169937,"cost_axis":42.15074720483203,"usd_per_1000_decisions":0.08478293103448274,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.28074855310842395,"p50_adjusted_s":0.7114971062168479,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"quyet-1-0-medium","name":"Quyet-1.0-Medium (Chinh Nguyen, Qwen3.5-4B decoder, option-letter logits)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/chinhnc/Quyet-1.0-Medium","listing":"ranked","ranked":true,"capability":59.24679183153445,"capability_eligible":true,"open_capability_rank":38,"composite":43.58925238992216,"intelligence":41.73350604436517,"calibration":76.76007761870373,"speed":89.46638461924755,"cost_axis":63.43457194636937,"usd_per_1000_decisions":0.01655189039408867,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): DeepInfra Qwen/Qwen3.5-4B list price $0.03/M input (the exact base weights; not on OpenRouter; same reference as reflex-4b, jobe and typecastlm), read 2026-09-24 x the package's own measured usage.input_tokens; one forward pass, nothing generated; run in process on 1x H100 80GB (Lium), quyet 1.0.0; estimated, not charged","p50_raw_s":0.07741166599589633,"p50_adjusted_s":0.3048233319917927,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"lev","name":"lev (Interfaze AI, Qwen3.5-4B + LoRA r32, label-token readout + candidate-path head, per-bucket temperatures)","board":"open","model_pin":"interfaze-ai/lev@7bdc748dffebd85b57ee0dbea8f994c6354fed31 + Abhinavexists/lev@cf104b69329302e4eac674a730c71f3511047db8","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/interfaze-ai/lev","listing":"ranked","ranked":true,"capability":61.70420936973613,"capability_eligible":true,"open_capability_rank":35,"composite":42.177760514319104,"intelligence":41.084702071234844,"calibration":82.32371666823741,"speed":88.00251499520694,"cost_axis":61.79891447724963,"usd_per_1000_decisions":0.01876590517241379,"price_kind":"estimate","price_basis":"latest published v1.5 row (v1.5.8 combined file) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3); self-hosted estimate, not measured hosted billing","p50_raw_s":0.10230199046782218,"p50_adjusted_s":0.3546039809356444,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"ninfer-qwen3.8-flash-next","name":"NInfer Qwen3.8-Flash-Next mixed","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/igorls/ninfer","listing":"ranked","ranked":true,"capability":69.67631278082561,"capability_eligible":false,"open_capability_rank":null,"composite":41.96785076957165,"intelligence":48.980301026876646,"calibration":90.37232453477458,"speed":89.33173525631216,"cost_axis":42.53362878447125,"usd_per_1000_decisions":0.08232764778325123,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.08002880541607738,"p50_adjusted_s":0.3100576108321548,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jev-local","name":"jev-local (Qwen3.5-9B)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/us/jev-local","listing":"ranked","ranked":true,"capability":57.458381676033426,"capability_eligible":true,"open_capability_rank":43,"composite":41.328564334669615,"intelligence":41.85041350559534,"calibration":73.06634984647151,"speed":75.55955847735143,"cost_axis":58.85554343293177,"usd_per_1000_decisions":0.023522413793103446,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.3653302099555731,"p50_adjusted_s":0.8806604199111462,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decider-4b-v2","name":"decider-4b v2 (Mapika)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Mapika/decider","listing":"ranked","ranked":true,"capability":64.88166453375648,"capability_eligible":true,"open_capability_rank":24,"composite":41.240894350857424,"intelligence":40.12597268917964,"calibration":89.63735637833331,"speed":91.78549329698745,"cost_axis":64.54376873042901,"usd_per_1000_decisions":0.015201077586206898,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.024963717442005873,"p50_adjusted_s":0.19992743488401174,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"classone-qwen3.5-9b","name":"ClassOne Qwen 3.5 9B (Qwen3.5-9B backbone with ClassOne decision heads, schema-constrained single forward pass, self-hosted)","board":"open","model_pin":"devops-thiago/classone-qwen3.5-9b@a45d16f1f5d1840a420f8ede59fa5cca3f079961","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/devops-thiago/classone-qwen3.5-9b","listing":"ranked","ranked":true,"capability":50.471173593724146,"capability_eligible":true,"open_capability_rank":54,"composite":38.54185769307265,"intelligence":42.225671849648506,"calibration":58.71667533779978,"speed":90.24509258883624,"cost_axis":49.33529996402755,"usd_per_1000_decisions":0.04884536222071767,"price_kind":"estimate","price_basis":"ESTIMATE (exact base-model market reference; self-hosted open weights, no per-token tariff of its own). THE ONLY ROW OF THIS RUN WHOSE RATE NEEDED NO JUDGEMENT AT ALL: Qwen/Qwen3.5-9B is in the frozen module's BASE_REFERENCES and pricing_v15.reference() resolves it straight through to qwen/qwen3.5-9b at USD 0.10 per 1M input and USD 0.15 per 1M output in the dated 25 Sep snapshot (DeepInfra's listing of the same base agrees at (0.10, 0.15)) x this row's own measured tokens. Estimated, not charged. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: the readout is a distribution over candidate labels from one forward pass and nothing is generated, so the row's cost rests on the input rate alone and any other decision is reversible by arithmetic from the raw.","p50_raw_s":0.05682604300091043,"p50_adjusted_s":0.2636520860018209,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jevk5-v0.3-4b","name":"JevK5 v0.3 (4B)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/alibiserikbay/JevK5","listing":"ranked","ranked":true,"capability":64.20716028123837,"capability_eligible":true,"open_capability_rank":25,"composite":37.36493548751909,"intelligence":38.53201254374527,"calibration":89.88230801873149,"speed":93.9472917853559,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.018816601019352674,"p50_adjusted_s":0.18763320203870534,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"metask-jev-4b","name":"metask-jev-4b","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/metask-ai/metask-jev","listing":"ranked","ranked":true,"capability":61.79110407115073,"capability_eligible":true,"open_capability_rank":33,"composite":37.34931073616377,"intelligence":39.166454612935595,"calibration":84.41575352936586,"speed":90.78239008886905,"cost_axis":57.7311327584882,"usd_per_1000_decisions":0.025642610837438424,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.05397228497895412,"p50_adjusted_s":0.25794456995790827,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"deck-4b-v1-0","name":"deck-4B v1.0 (krishna765, JevK5 + LoRA, FP8 weight-only)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://github.com/krishna-gogineni-765/deck","listing":"ranked","ranked":true,"capability":63.60791322392738,"capability_eligible":true,"open_capability_rank":27,"composite":37.26498528662829,"intelligence":38.62465958342338,"calibration":88.59116686443139,"speed":89.93848373337222,"cost_axis":63.46145558281372,"usd_per_1000_decisions":0.01651777251184834,"price_kind":"estimate","price_basis":"ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): DeepInfra Qwen/Qwen3.5-4B list price $0.03/M input, $0.15/M output (base of JevK5, the adapter's base checkpoint; not on OpenRouter; frozen 25 Sep snapshot per I-3, same reference as reflex-4b, jobe, typecastlm and Quyet) x the server's own measured usage.input_tokens; logit readout, 0 output tokens; FP8 weight-only on 1x RTX PRO 6000 96GB (Lium; v1.6 GPU-class deviation from source H100), deck-4b 1.0.0 + jevk5 f944fe37; estimated, not charged","p50_raw_s":0.07620866177603602,"p50_adjusted_s":0.30241732355207207,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"clef-flash","name":"Clef-Flash (Cloudflare, Qwen3.5-9B post-train with a joint schema head, multimodal, measured on text)","board":"open","model_pin":"Cloudflare/clef-flash@17f0b0ad64efb65d273590632833508766b2aae6","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/Cloudflare/clef-flash","listing":"ranked","ranked":true,"capability":63.824241207952895,"capability_eligible":true,"open_capability_rank":26,"composite":36.635050479655426,"intelligence":42.18368888957246,"calibration":85.46479352633332,"speed":88.4044196893156,"cost_axis":46.79963593205177,"usd_per_1000_decisions":0.05933971674876847,"price_kind":"estimate","price_basis":"latest published v1.5 row (v1.5.8 combined file) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied; self-hosted estimate, not measured hosted billing","p50_raw_s":0.10283017950132489,"p50_adjusted_s":0.3556603590026498,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"evalengine-decision-4b","name":"Decision-4B (Eval Engine / Chromia)","board":"open","model_pin":"evalengine/decision-4b-gguf@08859d1d8307f7b7ea452549e303b7805c35e350","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/evalengine/decision-4b-gguf","listing":"ranked","ranked":true,"capability":59.09507049015456,"capability_eligible":true,"open_capability_rank":39,"composite":35.43243109909669,"intelligence":37.93489123230218,"calibration":80.25524974800695,"speed":91.06043780349009,"cost_axis":65.87776871337948,"usd_per_1000_decisions":0.013721693349753695,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off, already marked deprecated at that cutoff (replaced by Qwen/Qwen3.5-9B); re-score if it changes (I-3). Frozen market reference deepinfra:Qwen/Qwen3.5-4B.","p50_raw_s":0.04748258483596146,"p50_adjusted_s":0.24496516967192292,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"janus-4b","name":"janus 4B (Icarus AI / cmxu, Qwen3.5-4B + LoRA r64 and pointer decision head)","board":"open","model_pin":"cmxu/janus-4b@fc86a8fbaa99de8dbac5aee46aaa56d1a853b876","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/cmxu/janus-4b","listing":"ranked","ranked":true,"capability":62.85256571162924,"capability_eligible":true,"open_capability_rank":28,"composite":35.18351981268845,"intelligence":37.57770128683752,"calibration":88.12743013642095,"speed":93.17471817821269,"cost_axis":64.41461114142905,"usd_per_1000_decisions":0.015352518472906404,"price_kind":"estimate","price_basis":"add-requests run 47 (v1.5 basis, unpublished) base-model price floor deepinfra:Qwen/Qwen3.5-4B","p50_raw_s":0.029892400838434696,"p50_adjusted_s":0.2097848016768694,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"qwen35-9b-jev-data-mix-v2","name":"Qwen3.5-9B Jev-like data-mix v2","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/jsaurabh/qwen3.5-9b-jev-data-mix-v2","listing":"ranked","ranked":true,"capability":60.86540021228136,"capability_eligible":false,"open_capability_rank":null,"composite":33.58966697347588,"intelligence":41.94184606895763,"calibration":79.78895435560509,"speed":85.40793837616167,"cost_axis":45.690218065856605,"usd_per_1000_decisions":0.06461391625615764,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.16828587953932583,"p50_adjusted_s":0.4865717590786517,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jevk5-v02","name":"JevK5 v0.2.0","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/allebee/jevk5","listing":"ranked","ranked":true,"capability":62.824112319658205,"capability_eligible":true,"open_capability_rank":29,"composite":32.09901620353128,"intelligence":36.261123811123895,"calibration":89.38710082819252,"speed":91.56570599843724,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.04039452798315324,"p50_adjusted_s":0.23078905596630647,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jevone","name":"JevOne","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/juspay/jev-one","listing":"ranked","ranked":true,"capability":68.8604478570867,"capability_eligible":false,"open_capability_rank":null,"composite":31.14510331640707,"intelligence":45.955077428728934,"calibration":91.76581828544445,"speed":89.83885080776159,"cost_axis":39.84212731591276,"usd_per_1000_decisions":0.10121908866995075,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.06179859535768628,"p50_adjusted_s":0.2735971907153726,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kahn1","name":"Kahn1 4B (Okura66, Qwen3.5-4B LoRA merge, author's sysone engine)","board":"open","model_pin":"Okura66/Kahn1-Qwen3.5-4B@f3f9910f995b0e345a6e8a9cb0aeed1e06bafaf6","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/Okura66/Kahn1-Qwen3.5-4B","listing":"ranked","ranked":true,"capability":61.74359439203453,"capability_eligible":true,"open_capability_rank":34,"composite":31.117574813020727,"intelligence":39.71971801341685,"calibration":83.76747077065221,"speed":89.97531755002697,"cost_axis":46.37729040230682,"usd_per_1000_decisions":0.06129480295566502,"price_kind":"estimate","price_basis":"ESTIMATE, ours (the author's server returns no token usage; Florian 7 Oct 2026: use our own estimate): 829,523 input tokens counted with the author's tokenizer and prompt builder over the 1,624 v1.5 items x n_permutations 3 (the author's server default) = 2,488,569 tokens / 1,624 decisions x USD 0.04 per 1M input (deepinfra:Qwen/Qwen3.5-4B, frozen 25 Sep snapshot), 0 output tokens = USD 0.061295 per 1,000 decisions; carried like every Qwen3.5-4B row on this board. Cross-check: the same count on the v1.6 common cost basis (1,477 items, x3) gives USD 0.0597 per 1,000 (within 3 %)","p50_raw_s":0.06609634149936028,"p50_adjusted_s":0.2821926829987206,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"typecastlm-1.4.0","name":"TypeCastLM 1.4.0 (Mikhail Gribov, Qwen3.5-4B computed head)","board":"open","model_pin":"mihailgribov/typecastlm-qwen3.5-3.8b@0984c3e832f7c28495134bc1d22c66bb0e0a555d","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/mihailgribov/typecastlm-qwen3.5-3.8b","listing":"ranked","ranked":true,"capability":56.39414821891969,"capability_eligible":true,"open_capability_rank":49,"composite":29.827386072784904,"intelligence":35.43137209354293,"calibration":77.35692434429644,"speed":92.95077486551261,"cost_axis":64.79986008365586,"usd_per_1000_decisions":0.014905206499661474,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights). NO JUDGEMENT OF OURS: pricing_v15.reference() resolves Qwen/Qwen3.5-4B straight through to deepinfra:Qwen/Qwen3.5-4B at USD 0.03 per 1M input and USD 0.15 per 1M output in the frozen 25 Sep snapshot (the same rate aplomb-1 carries for the same base) x this row's own measured tokens. Output tokens are 0 in every answered record: one prefill, option-label readout, nothing generated. Estimated, not charged.","p50_raw_s":0.030360452496097423,"p50_adjusted_s":0.21072090499219484,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"hopper","name":"Hopper","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/HopitAI/hopper","listing":"ranked","ranked":true,"capability":62.40936838189898,"capability_eligible":true,"open_capability_rank":30,"composite":28.852652318563656,"intelligence":34.73970870443837,"calibration":90.07902805935959,"speed":91.11434937226889,"cost_axis":62.25627699896333,"usd_per_1000_decisions":0.018118577586206896,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.0564777513500303,"p50_adjusted_s":0.26295550270006063,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"imajev-4b-rtx5090-a2","name":"Imajev-4B (RTX 5090)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/fstandhartinger/jevbench/issues/110","listing":"ranked","ranked":true,"capability":61.89494846465412,"capability_eligible":true,"open_capability_rank":31,"composite":28.655751132526984,"intelligence":34.62714018309005,"calibration":89.1627567462182,"speed":90.51878000442085,"cost_axis":63.26458220826236,"usd_per_1000_decisions":0.016769261083743846,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (I-2): measured input tokens; zero generated output tokens for signed logits readout. M2 floor uses the 25 Sep 2026 DeepInfra Qwen3.5-4B snapshot rates (USD 0.03/M input, USD 0.15/M output); the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen3.5-9B.","p50_raw_s":0.07120202149963006,"p50_adjusted_s":0.29240404299926015,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"malkuth-4b","name":"Malkuth-4B (newfull5, Kev post-train)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/newfull5/malkuth","listing":"ranked","ranked":true,"capability":61.139059860562405,"capability_eligible":true,"open_capability_rank":36,"composite":26.197476126697616,"intelligence":33.840703805987175,"calibration":88.43741591513763,"speed":89.51571180654442,"cost_axis":55.82433433950372,"usd_per_1000_decisions":0.02968398399014778,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.07427763848681934,"p50_adjusted_s":0.2985552769736387,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jqv","name":"jqv (Qwen3-32B zero-shot)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Octalab-Inc/jqv","listing":"ranked","ranked":true,"capability":61.8647697220168,"capability_eligible":true,"open_capability_rank":32,"composite":25.391439220985372,"intelligence":33.85815624201783,"calibration":89.87138320201578,"speed":83.1032423977654,"cost_axis":51.172592607797476,"usd_per_1000_decisions":0.042420935960591134,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.09740161104127765,"p50_adjusted_s":0.3448032220825553,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jpt-35b-a3b","name":"JPT-35B-A3B (Qwen3.5-35B-A3B fine-tune)","board":"open","model_pin":"kirp/jpt-35b-a3b@f9b45a14df4f7eaf5551f1ab26607237604a8588","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/kirp/jpt-35b-a3b","listing":"ranked","ranked":true,"capability":70.67673257774709,"capability_eligible":false,"open_capability_rank":null,"composite":25.357657195103958,"intelligence":51.42208418374838,"calibration":89.9313809717458,"speed":90.51950606084544,"cost_axis":33.627578514713974,"usd_per_1000_decisions":0.16308501184834123,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): OpenRouter qwen/qwen3.5-35b-a3b in the frozen 25 Sep snapshot (exact declared base; no explicit listing-decision line in the frozen module: release-lane line), USD 0.3125 per 1M input and USD 1.25 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.070619512029225,"p50_adjusted_s":0.29123902405845004,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decider-35b-a3b","name":"decider-35b-a3b (Mapika)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Mapika/decider-35b-a3b","listing":"ranked","ranked":true,"capability":66.1715118587882,"capability_eligible":false,"open_capability_rank":null,"composite":24.772887446577812,"intelligence":48.72471958402083,"calibration":83.61830413355554,"speed":91.27357492980725,"cost_axis":34.38579876859529,"usd_per_1000_decisions":0.1538650708128079,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.04968994203954935,"p50_adjusted_s":0.2493798840790987,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"coco-decision-4b","name":"CoCo-Decision-4B (corners-ai, LoRA on Qwen3.5-4B, option-label logits of one pass, served by oh-my-jev)","board":"open","model_pin":"corners-ai/CoCo-Decision-4B-Ko@b580ede993493690212a07ed0e6cb972868cc808","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/corners-ai/CoCo-Decision-4B-Ko","listing":"ranked","ranked":true,"capability":56.92005372388988,"capability_eligible":true,"open_capability_rank":46,"composite":24.64791535308594,"intelligence":32.65682750200105,"calibration":81.18327994577871,"speed":91.3615355688411,"cost_axis":65.17140064644872,"usd_per_1000_decisions":0.014486161137440754,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights). NO JUDGEMENT OF OURS: pricing_v15.reference() resolves Qwen/Qwen3.5-4B straight through to deepinfra:Qwen/Qwen3.5-4B at USD 0.03 per 1M input and USD 0.15 per 1M output in the frozen 25 Sep snapshot (the same rate aplomb-1 carries for the same base) x this row's own measured tokens. Output tokens are 0 in every answered record: one prefill, option-label readout, nothing generated. Estimated, not charged.","p50_raw_s":0.044485659993370064,"p50_adjusted_s":0.23897131998674012,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"manchego-v21-v15","name":"Manchego v2.1","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://github.com/nschlaepfer/manchego-serve","listing":"ranked","ranked":true,"capability":58.27541943544253,"capability_eligible":true,"open_capability_rank":40,"composite":24.487333472500133,"intelligence":32.54303915979197,"calibration":84.0077997110931,"speed":90.7347510499745,"cost_axis":64.32996474661243,"usd_per_1000_decisions":0.015452586206896551,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: frozen 25 Sep DeepInfra Qwen/Qwen3.5-4B market reference, already marked deprecated at that cutoff (replaced by Qwen/Qwen3.5-9B). $0.03/M input, zero generated output; 836,500 measured input tokens across 1,624 decisions. Frozen v1.5 base-model floor applied; no bookable Manchego tariff claimed. Re-score if the basis changes.","p50_raw_s":0.05371422949974658,"p50_adjusted_s":0.2574284589994932,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decision-4b-v12","name":"Decision 4B v1.2 (FlyMyJev, Qwen3.5-4B + LoRA)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://benchmarkheaven.com/jev-models/decision-4b-v12","listing":"ranked","ranked":true,"capability":60.41399746452394,"capability_eligible":true,"open_capability_rank":37,"composite":24.463556699266523,"intelligence":32.40310972936693,"calibration":88.42488519968094,"speed":93.92936299294371,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.019149313447996974,"p50_adjusted_s":0.18829862689599394,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"raw-qwen3-4b-instruct-2507","name":"Raw Qwen3 4B Instruct 2507 direct logits","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507","listing":"ranked","ranked":true,"capability":34.56646466717713,"capability_eligible":true,"open_capability_rank":80,"composite":21.792436038282993,"intelligence":33.95117742100547,"calibration":35.18175191334879,"speed":91.16296741568695,"cost_axis":63.36068112101963,"usd_per_1000_decisions":0.01664602832512315,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.04603758594021201,"p50_adjusted_s":0.24207517188042402,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jev-27b-autotrust","name":"JEV-27B (AutoTrust, Qwen3.8-27B)","board":"open","model_pin":"autotrust/JEV-27B@51740a8891c2a8baefd969237fd44187b3e3a115","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/autotrust/JEV-27B","listing":"ranked","ranked":true,"capability":73.86417008485257,"capability_eligible":false,"open_capability_rank":null,"composite":21.291230595611577,"intelligence":56.828664559360675,"calibration":90.89967561034447,"speed":86.89503266505372,"cost_axis":31.02571003473804,"usd_per_1000_decisions":0.19913289099526063,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.1052485710097244,"p50_adjusted_s":0.3604971420194488,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"torchcast-decision-27b","name":"torchcast-decision-27b (Torchcast AI, Qwen3.8-27B fine-tune)","board":"open","model_pin":"torchcast-ai/torchcast-decision-27b@08645834899cb8e00d370b77a3f4e0baa3241dd1","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/torchcast-ai/torchcast-decision-27b","listing":"ranked","ranked":true,"capability":81.56565836114741,"capability_eligible":false,"open_capability_rank":null,"composite":21.273291266043532,"intelligence":71.75187578095783,"calibration":91.379440941337,"speed":87.72436167205926,"cost_axis":30.341044859222123,"usd_per_1000_decisions":0.20987715639810428,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.09728601449751295,"p50_adjusted_s":0.3445720289950259,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"nimble-9b","name":"Bespoke Nimble 9B (Bespoke Labs)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/bespokelabsai/nimble","listing":"ranked","ranked":true,"capability":56.514998578191054,"capability_eligible":false,"open_capability_rank":null,"composite":21.082616704957896,"intelligence":43.14237010986392,"calibration":69.88762704651819,"speed":86.11327010751721,"cost_axis":36.75037423027614,"usd_per_1000_decisions":0.1283278325123153,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.14493575389496982,"p50_adjusted_s":0.43987150778993966,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"reflex-4b","name":"reflex 4B (kshetrajna12)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/kshetrajna12/reflex","listing":"ranked","ranked":true,"capability":58.95740703790636,"capability_eligible":false,"open_capability_rank":null,"composite":20.637311182702398,"intelligence":30.935558296483705,"calibration":86.97925577932901,"speed":68.7900017304639,"cost_axis":63.142712976263866,"usd_per_1000_decisions":0.016926853448275863,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":1.3577107690507546,"p50_adjusted_s":2.865421538101509,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"pplx-decider-v1-1-27b","name":"Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)","board":"open","model_pin":"perplexity-ai/pplx-decider-v1.1-27b@3b45dead91dfa6d95aad6b95764a606fab2bf7a6","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/perplexity-ai/pplx-decider-v1.1-27b","listing":"ranked","ranked":true,"capability":82.79186839726479,"capability_eligible":false,"open_capability_rank":null,"composite":20.617642759667973,"intelligence":75.71677945512734,"calibration":89.86695733940225,"speed":86.79279341461623,"cost_axis":29.886617009940878,"usd_per_1000_decisions":0.21732654028436021,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.12236087050405331,"p50_adjusted_s":0.39472174100810664,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jade","name":"JADE (Qwen3.8-27B LoRA)","board":"open","model_pin":"theunnecessarythings/JADE@8e9f9a3aa4293190fcc4a6ff17ff5f3251f4c270","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/theunnecessarythings/JADE","listing":"ranked","ranked":true,"capability":49.24946957396523,"capability_eligible":false,"open_capability_rank":null,"composite":20.194459825016597,"intelligence":47.93439876939806,"calibration":50.5645403785324,"speed":85.53222225487117,"cost_axis":33.57969498589053,"usd_per_1000_decisions":0.16368548408937036,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.12725698550639208,"p50_adjusted_s":0.4045139710127842,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"typecastlm","name":"typecastlm (Mikhail Gribov, Qwen3.5-4B computed head)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/mihail-gribov/typecastlm","listing":"ranked","ranked":true,"capability":56.544258540063545,"capability_eligible":true,"open_capability_rank":48,"composite":19.72256430004839,"intelligence":29.78900656204624,"calibration":83.29951051808085,"speed":92.76354061232337,"cost_axis":63.95900966831593,"usd_per_1000_decisions":0.01589887315270936,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.02951649052556604,"p50_adjusted_s":0.20903298105113208,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jebadiah-27b","name":"Jebadiah 27B (Frontier Infra, Qwen3.8-27B LoRA)","board":"open","model_pin":"frontier-infra/jebadiah-27b@4d3ed81640bbc015b47976bedae67212a5fb39cd","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/frontier-infra/jebadiah-27b","listing":"ranked","ranked":true,"capability":70.3119788331729,"capability_eligible":false,"open_capability_rank":null,"composite":19.255307672764186,"intelligence":57.58551644935426,"calibration":83.03844121699153,"speed":85.83700487110255,"cost_axis":29.935481526523958,"usd_per_1000_decisions":0.2165129857819905,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.13926563448330853,"p50_adjusted_s":0.4285312689666171,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decision-2-0-vega-27b","name":"Decision 2.0 Vega 27B (vLLM-SR, Qwen3.8-27B)","board":"open","model_pin":"vllm-sr/Decision-2.0-Vega-27B@7aec49ae11a18741706da549ab626b9052795fe7","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/vllm-sr/Decision-2.0-Vega-27B","listing":"ranked","ranked":true,"capability":71.30534158412956,"capability_eligible":false,"open_capability_rank":null,"composite":18.883756282485887,"intelligence":60.4093162408047,"calibration":82.20136692745442,"speed":83.80610555916236,"cost_axis":29.632311028770005,"usd_per_1000_decisions":0.22161014218009478,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.1901585105806589,"p50_adjusted_s":0.5303170211613178,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decision-4b-v11","name":"Decision 4B v1.1 (FlyMyJev, Qwen3.5-4B + LoRA)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://benchmarkheaven.com/jev-models/decision-4b-v11","listing":"ranked","ranked":true,"capability":57.18675493386165,"capability_eligible":true,"open_capability_rank":44,"composite":18.718356274892123,"intelligence":29.134168280172865,"calibration":85.23934158755043,"speed":93.95641941218838,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.019026647554710507,"p50_adjusted_s":0.188053295109421,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"autojev-27b","name":"AutoJev-27B (denis-pplx, Qwen3.8-27B)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/denis-pplx/autojev-27b","listing":"ranked","ranked":true,"capability":71.61983491755967,"capability_eligible":false,"open_capability_rank":null,"composite":18.463427917707932,"intelligence":55.62422706919757,"calibration":87.61544276592178,"speed":88.60757446254297,"cost_axis":29.365937020749882,"usd_per_1000_decisions":0.22618758620689655,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.09737715951632708,"p50_adjusted_s":0.3447543190326542,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"eikos-27b","name":"Eikos-27B (caiovicentino1, Qwen3.8-27B)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/caiovicentino1/eikos-27b","listing":"ranked","ranked":true,"capability":77.18992599407068,"capability_eligible":false,"open_capability_rank":null,"composite":18.077889022432696,"intelligence":64.40270761245412,"calibration":89.97714437568723,"speed":88.08630815059152,"cost_axis":28.689752882469676,"usd_per_1000_decisions":0.23823646551724137,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.11261826660484076,"p50_adjusted_s":0.37523653320968153,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"ninfer-qwen3.8-27b","name":"NInfer Qwen3.8-27B NVFP4","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/igorls/ninfer","listing":"ranked","ranked":true,"capability":67.99284998901135,"capability_eligible":false,"open_capability_rank":null,"composite":17.71600496541158,"intelligence":51.800126951055375,"calibration":84.18557302696732,"speed":90.47734239273063,"cost_axis":29.118887844204664,"usd_per_1000_decisions":0.2305174137931034,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.05501334450673312,"p50_adjusted_s":0.26002668901346626,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"openjev-thinking","name":"OpenJev (thinking, BF16)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/razorback16/openjev","listing":"ranked","ranked":true,"capability":75.87983429229429,"capability_eligible":false,"open_capability_rank":null,"composite":17.265218966202397,"intelligence":70.34621384425424,"calibration":81.41345474033434,"speed":72.55270827201397,"cost_axis":28.51480924396573,"usd_per_1000_decisions":0.24145693965517237,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.8569832294015214,"p50_adjusted_s":1.8639664588030427,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-jev-zefan-9b","name":"Open-Jev 9B (Zefan Cai)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Zefan-Cai/Open-Jev","listing":"ranked","ranked":true,"capability":61.84549459243814,"capability_eligible":false,"open_capability_rank":null,"composite":17.045611746249723,"intelligence":43.88591848753404,"calibration":79.80507069734225,"speed":74.39486349899005,"cost_axis":33.05490478272432,"usd_per_1000_decisions":0.17041317733990147,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.6175056385109201,"p50_adjusted_s":1.3850112770218401,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"standardone-8b","name":"Standard One 8B (Standard Thinking)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/StandardThinking/StandardOne-8B","listing":"ranked","ranked":true,"capability":58.40967348707115,"capability_eligible":false,"open_capability_rank":null,"composite":16.917108299009165,"intelligence":32.80426604507879,"calibration":84.01508092906352,"speed":92.76442849252426,"cost_axis":43.283250331904235,"usd_per_1000_decisions":0.07772456896551724,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.03173699742183089,"p50_adjusted_s":0.21347399484366178,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"clef","name":"Clef (Cloudflare, Qwen3.8-27B post-train with a joint schema head, multimodal, measured on text)","board":"open","model_pin":"Cloudflare/clef@2f3de3dd85f379784083b0814d997ab627200f0c","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/Cloudflare/clef","listing":"ranked","ranked":true,"capability":75.32934491938597,"capability_eligible":false,"open_capability_rank":null,"composite":16.753692861922005,"intelligence":59.08165501447566,"calibration":91.57703482429629,"speed":83.3998209667487,"cost_axis":28.102157220114762,"usd_per_1000_decisions":0.24922681034482758,"price_kind":"estimate","price_basis":"latest published v1.5 row (v1.5.8 combined file) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied; self-hosted estimate, not measured hosted billing","p50_raw_s":0.21318734902888536,"p50_adjusted_s":0.5763746980577708,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jev-qwen3.5-9b-base-nvfp4","name":"JEV Qwen3.5-9B Base NVFP4","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/WIlfLin/JEV-Qwen3.5-9B-Base-NVFP4","listing":"ranked","ranked":true,"capability":57.14200691217819,"capability_eligible":true,"open_capability_rank":45,"composite":15.97555620862577,"intelligence":29.251131763478327,"calibration":85.03288206087805,"speed":93.62779491136789,"cost_axis":47.592235403520306,"usd_per_1000_decisions":0.05583743842364531,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (proxy tokens, I-2): exact base-model market reference","p50_raw_s":0.02267279999796301,"p50_adjusted_s":0.19534559999592602,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"semif-qwen3.5-4b","name":"SemIf, formerly OpenJev (Qwen3.5-4B, TheoLeeCJ)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/TheoLeeCJ/openjev","listing":"ranked","ranked":true,"capability":55.35036926555066,"capability_eligible":true,"open_capability_rank":50,"composite":15.517179498801362,"intelligence":27.10113669038628,"calibration":83.59960184071504,"speed":90.76560119421038,"cost_axis":63.07359395281681,"usd_per_1000_decisions":0.01701689039408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.05204228049842641,"p50_adjusted_s":0.25408456099685284,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"bev-bonsai-27b","name":"Bev / Bonsai 27B","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/Reza2kn/Bev","listing":"ranked","ranked":true,"capability":66.13873250860377,"capability_eligible":false,"open_capability_rank":null,"composite":11.7393704396132,"intelligence":43.77513601921869,"calibration":88.50232899798883,"speed":72.98222461463935,"cost_axis":28.231834531826323,"usd_per_1000_decisions":0.2467585344827586,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: documented hosted-model estimate. Frozen market reference qwen/qwen3.8-27b.","p50_raw_s":0.9861058448441327,"p50_adjusted_s":2.1222116896882652,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"litjev","name":"LitJev (Qwen3.8-27B)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/zhengxuyu/litjev","listing":"ranked","ranked":true,"capability":66.60329337630368,"capability_eligible":false,"open_capability_rank":null,"composite":11.458886170296882,"intelligence":43.28549422942267,"calibration":89.9210925231847,"speed":68.17671225113584,"cost_axis":28.35809912639418,"usd_per_1000_decisions":0.2443787068965517,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":1.4730074456892908,"p50_adjusted_s":3.0960148913785814,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"localjev-qwen3.5-4b","name":"local-jev Qwen3.5-4B","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/amithgc/local-jev","listing":"ranked","ranked":true,"capability":54.99411342282565,"capability_eligible":true,"open_capability_rank":51,"composite":11.438161297684031,"intelligence":24.145363218466134,"calibration":85.84286362718517,"speed":86.10068849774444,"cost_axis":59.269627658320445,"usd_per_1000_decisions":0.022786576354679802,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.1226705745793879,"p50_adjusted_s":0.39534114915877583,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"system-one-sg","name":"system-one (Qwen3-8B, Sean Goedecke)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/sgoedecke/system-one","listing":"ranked","ranked":true,"capability":29.718120999285617,"capability_eligible":false,"open_capability_rank":null,"composite":11.027723483913586,"intelligence":29.271508779158296,"calibration":30.16473321941294,"speed":90.8867694864598,"cost_axis":44.96855803039237,"usd_per_1000_decisions":0.06829381896551724,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.04996280049090274,"p50_adjusted_s":0.24992560098180547,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kev-8b","name":"kev 8B (research preview)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/jaredpalmer/kev","listing":"ranked","ranked":true,"capability":50.0091761607623,"capability_eligible":false,"open_capability_rank":null,"composite":10.62283914209564,"intelligence":29.26217941280341,"calibration":70.7561729087212,"speed":88.97143473610706,"cost_axis":40.42054386073457,"usd_per_1000_decisions":0.09682374384236453,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.07448393944650888,"p50_adjusted_s":0.2989678788930178,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decision-2b","name":"Decision 2B (FlyMy.AI, v59)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://huggingface.co/flymy-ai/decision-2b-preview","listing":"ranked","ranked":true,"capability":54.82290333495299,"capability_eligible":true,"open_capability_rank":52,"composite":10.289708493744515,"intelligence":22.875473256337628,"calibration":86.77033341356835,"speed":90.39462801231525,"cost_axis":66.37275756963365,"usd_per_1000_decisions":0.013210160098522166,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.07514283142518252,"p50_adjusted_s":0.30028566285036506,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"opensourcejev-qwen35-4b-q4km","name":"OpenSourceJev (Qwen3.5-4B Q4_K_M, native llama.cpp)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/sabeel111/OpenSourceJev","listing":"ranked","ranked":true,"capability":48.571422117188824,"capability_eligible":true,"open_capability_rank":57,"composite":10.263493713193125,"intelligence":23.217995107845883,"calibration":73.92484912653177,"speed":77.06362388237805,"cost_axis":69.13657649738097,"usd_per_1000_decisions":0.0106851539408867,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.29475566232576966,"p50_adjusted_s":0.7395113246515393,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-alternative-jev","name":"open-alternative-jev (Qwen3.5-4B, IkerMoel)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/ikermoel/open-alternative-jev","listing":"ranked","ranked":true,"capability":49.14497164755345,"capability_eligible":true,"open_capability_rank":56,"composite":9.435692571873393,"intelligence":22.34610151162883,"calibration":75.94384178347806,"speed":91.30216983336405,"cost_axis":63.27859776193043,"usd_per_1000_decisions":0.0167512315270936,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (proxy tokens, I-2): exact base-model market reference; reference deepinfra:Qwen/Qwen3.5-4B frozen at the 25 Sep cut-off; the snapshot records deprecation on 11 Jun 2026 and replacement by Qwen/Qwen3.5-9B; the frozen snapshot rates remain in use (I-3)","p50_raw_s":0.04333395400317386,"p50_adjusted_s":0.2366679080063477,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"raw-qwen3-8b","name":"Raw Qwen3 8B direct logits","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Qwen/Qwen3-8B","listing":"ranked","ranked":true,"capability":30.13533882926388,"capability_eligible":false,"open_capability_rank":null,"composite":9.30125972789427,"intelligence":26.54912920590792,"calibration":33.721548452619835,"speed":88.65282990878164,"cost_axis":45.53515592313027,"usd_per_1000_decisions":0.0653875104679803,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.07753071398474276,"p50_adjusted_s":0.30506142796948554,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decider-2b","name":"decider-2b (Mapika)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Mapika/decider-2b","listing":"ranked","ranked":true,"capability":44.58947913615145,"capability_eligible":true,"open_capability_rank":62,"composite":7.6704332123718,"intelligence":20.654576369710338,"calibration":68.52438190259257,"speed":94.55945724666857,"cost_axis":64.91766177740826,"usd_per_1000_decisions":0.014771046798029557,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.014745104941539466,"p50_adjusted_s":0.17949020988307893,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"nemotron-diffusion-8b","name":"Nemotron Diffusion 8B (pst2154, optimized vLLM)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://github.com/pst2154/Nemotron_Jev","listing":"ranked","ranked":true,"capability":47.46294541124625,"capability_eligible":true,"open_capability_rank":60,"composite":7.29748476516348,"intelligence":20.38848344571935,"calibration":74.53740737677316,"speed":93.8305724848098,"cost_axis":55.49170925809556,"usd_per_1000_decisions":0.03045157019704434,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.02195067796856165,"p50_adjusted_s":0.1939013559371233,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kev-4b","name":"kev 4B (research preview)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/jaredpalmer/kev","listing":"ranked","ranked":true,"capability":43.98178354242468,"capability_eligible":true,"open_capability_rank":64,"composite":6.611791819903415,"intelligence":19.510707720813777,"calibration":68.45285936403559,"speed":90.46585183268016,"cost_axis":65.78036497095258,"usd_per_1000_decisions":0.013824661330049262,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.07199650770053267,"p50_adjusted_s":0.29399301540106537,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kev-27b","name":"Kev 27B (Kev 1.0, Qwen3.8-27B)","board":"open","model_pin":"jaredpalmer/kev-27b@af0e6d551bdc2cc724f3e9d7a8bee1cd4fb8f7bf","last_measured_on":"2026-10-06","source_url":"https://huggingface.co/jaredpalmer/kev-27b","listing":"ranked","ranked":true,"capability":78.93896122956639,"capability_eligible":false,"open_capability_rank":null,"composite":5.77223551264444,"intelligence":64.16071886717361,"calibration":93.71720359195915,"speed":83.65684843680415,"cost_axis":18.302044660887503,"usd_per_1000_decisions":0.5287680162491537,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.18013913399772719,"p50_adjusted_s":0.5102782679954544,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"malkuth-2b","name":"Malkuth-2B (newfull5, Kev post-train)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/newfull5/malkuth","listing":"ranked","ranked":true,"capability":47.633125198824516,"capability_eligible":true,"open_capability_rank":59,"composite":3.992169201455145,"intelligence":15.917395195670291,"calibration":79.34885520197874,"speed":92.03462875001455,"cost_axis":65.5661191189412,"usd_per_1000_decisions":0.014053873152709357,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.04305429550004192,"p50_adjusted_s":0.23610859100008383,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"qwen3-reranker-4b","name":"Qwen3-Reranker-4B","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Qwen/Qwen3-Reranker-4B","listing":"ranked","ranked":true,"capability":44.7457252992122,"capability_eligible":true,"open_capability_rank":61,"composite":3.715006798009075,"intelligence":16.337944014530347,"calibration":73.15350658389404,"speed":82.14971164581814,"cost_axis":48.39770052432463,"usd_per_1000_decisions":0.05249000923645321,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: hosted exact-model reference (deepinfra:Qwen/Qwen3-Reranker-4B); no exact base-model floor applies","p50_raw_s":0.20409131213091314,"p50_adjusted_s":0.5581826242618263,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"simplejev-qwen3.8-27b-selfhost","name":"SimpleJev Qwen3.8-27B (self-hosted, v1.6 pool)","board":"open","model_pin":"Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 + featherless-ai/simple-jev@9c11582d3e631e0b118f3f3132ebbb00c62020c5","last_measured_on":"2026-10-06","source_url":"https://github.com/featherless-ai/simple-jev","listing":"ranked","ranked":true,"capability":74.89270871895144,"capability_eligible":false,"open_capability_rank":null,"composite":3.394848880196433,"intelligence":57.92931033189586,"calibration":91.85610710600702,"speed":76.19964460590472,"cost_axis":15.10193161538902,"usd_per_1000_decisions":0.6759827488151658,"price_kind":"estimate","price_basis":"ESTIMATE (base-model reference; self-hosted open weights, no public tariff): frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price, the pair the published autojev-27b, eikos-27b and open-jev-zefan-27b-v1.1 rows use, USD 0.42 per 1M input and USD 3.0 per 1M output x this row's own measured tokens. Estimated, not charged","p50_raw_s":0.6217738870036555,"p50_adjusted_s":1.393547774007311,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"zerank-2","name":"ZeroEntropy zerank-2","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/zeroentropy/zerank-2-reranker","listing":"ranked","ranked":true,"capability":49.39850940998648,"capability_eligible":true,"open_capability_rank":55,"composite":3.349795360318651,"intelligence":15.596611891228248,"calibration":83.20040692874471,"speed":82.89783918855927,"cost_axis":48.39770052432463,"usd_per_1000_decisions":0.05249000923645321,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: size-class proxy (deepinfra:Qwen/Qwen3-Reranker-4B); no exact base-model floor applies","p50_raw_s":0.17457342706620693,"p50_adjusted_s":0.4991468541324139,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"exaone-4.0-1.2b-jev-v0.3","name":"EXAONE-4.0-1.2B-JEV v0.3 (unofficial fine-tune of LGAI-EXAONE/EXAONE-4.0-1.2B, native option-label softmax, two reads averaged, self-hosted)","board":"open","model_pin":"carrtesy/EXAONE-4.0-1.2B-JEV@v0.3","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/carrtesy/EXAONE-4.0-1.2B-JEV","listing":"ranked","ranked":true,"capability":47.67995157039467,"capability_eligible":true,"open_capability_rank":58,"composite":3.266776278413127,"intelligence":14.831288392065307,"calibration":80.52861474872404,"speed":94.09940251974388,"cost_axis":57.919479713067595,"usd_per_1000_decisions":0.025274583615436694,"price_kind":"estimate","price_basis":"ESTIMATE (size-class reference; self-hosted open weights, no bookable listing for this base). THE RATE IS RUN 48'S, CARRIED UNCHANGED FOR THE SAME BASE: run 48 priced exaone-4.0-1.2b-jev (v0.2, identical base) at USD 0.02 per 1M input and 0 output, as the HIGHER of the two bracketing published rates, chosen in that direction because the rate was fixed after the measurement existed. Keeping it makes v0.3 directly comparable with v0.2 and takes no new decision. The frozen 25 Sep snapshot still contains NO listing for LGAI-EXAONE/EXAONE-4.0-1.2B (pricing_v15.reference() raises: 'base model lacks an explicit market-listing decision'), and still brackets 1.2B between deepinfra:Qwen/Qwen3.5-0.8B at (0.01, 0.05) and deepinfra:Qwen/Qwen3.5-2B at (0.02, 0.10). Estimated, not charged. RELEASE-LANE LINE, recorded and not acted on, AND IT IS RUN 48'S CARRIED LINE, NOT A NEW ONE: LGAI-EXAONE/EXAONE-4.0-1.2B must be adopted into the frozen UNLISTED_BASE_MODELS table before either EXAONE row may be published. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: the readout is a distribution over candidate labels from one forward pass and nothing is generated, so the row's cost rests on the input rate alone and any other decision is reversible by arithmetic from the raw.","p50_raw_s":0.020266480001737364,"p50_adjusted_s":0.19053296000347472,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-jev-zefan-2b","name":"Open-Jev 2B (Zefan Cai)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Zefan-Cai/Open-Jev","listing":"ranked","ranked":true,"capability":47.71137138509946,"capability_eligible":false,"open_capability_rank":null,"composite":3.2383990318693976,"intelligence":21.804441188912403,"calibration":73.61830158128652,"speed":77.13731150142542,"cost_axis":33.05490478272432,"usd_per_1000_decisions":0.17041317733990147,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":0.43650909198913723,"p50_adjusted_s":1.0230181839782744,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-jev-zefan-27b-v1.1","name":"Open-Jev 27B v1.1 (Zefan Cai, LoRA + decision head on Qwen3.8-27B)","board":"open","model_pin":"HF ZefanCai/Open-Jev-27B-v1.1@28cf73067d5b337860bbef3c85b8b82ba8730956; base Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0; author loader Zefan-Cai/Open-Jev@3308a15ccd7eea1df7a37d6ddc39b023b801ba16","last_measured_on":"2026-10-04","source_url":"https://huggingface.co/ZefanCai/Open-Jev-27B-v1.1","listing":"ranked","ranked":true,"capability":67.96306879751505,"capability_eligible":false,"open_capability_rank":null,"composite":2.8808671915746493,"intelligence":52.45777412397041,"calibration":83.46836347105969,"speed":72.50871441156775,"cost_axis":14.357426070787312,"usd_per_1000_decisions":0.7157353448275862,"price_kind":"estimate","price_basis":"latest published v1.5 row (v1.5.8 combined file) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): Frozen BASE_REFERENCES Qwen/Qwen3.8-27B market price (USD 0.42/M input, 3.00/M output) applied to 2,767,510 input tokens and zero output tokens over 1,624 decisions; self-hosted estimate, not measured hosted billing. Lineage verified by architecture; no pricing gate. The author method encodes each candidate independently.","p50_raw_s":0.7902996460907161,"p50_adjusted_s":1.7305992921814322,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"raw-phi-4-mini","name":"Raw Phi-4 mini direct logits","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/microsoft/Phi-4-mini-instruct","listing":"ranked","ranked":true,"capability":35.33785192837895,"capability_eligible":true,"open_capability_rank":79,"composite":2.4794266829120515,"intelligence":13.660237978082595,"calibration":57.015465878675315,"speed":91.89255362835293,"cost_axis":53.22051180195924,"usd_per_1000_decisions":0.03625068965517241,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.04178677382878959,"p50_adjusted_s":0.23357354765757918,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"simplejev-qwen3.5-0.8b","name":"SimpleJev (Qwen3.5-0.8B, CPU)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/featherless-ai/simple-jev","listing":"ranked","ranked":true,"capability":31.59915405528791,"capability_eligible":false,"open_capability_rank":null,"composite":2.112084174105171,"intelligence":13.000426205687345,"calibration":50.197881904888476,"speed":58.929227255635794,"cost_axis":70.31175519898045,"usd_per_1000_decisions":0.009763559113300492,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":4.323511580238119,"p50_adjusted_s":8.797023160476238,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"smalljev","name":"smalljev semantic-v9","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/isHeSatoshi/smalljev","listing":"ranked","ranked":true,"capability":42.95073007048671,"capability_eligible":true,"open_capability_rank":68,"composite":0.7719455446134506,"intelligence":8.670533657732655,"calibration":77.23092648324078,"speed":90.21131723708834,"cost_axis":60.77469204714053,"usd_per_1000_decisions":0.020300665024630543,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.06657194206491113,"p50_adjusted_s":0.2831438841298223,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kev-0.6b","name":"kev 0.6B (research preview)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/jaredpalmer/kev","listing":"ranked","ranked":true,"capability":36.93267237588606,"capability_eligible":true,"open_capability_rank":77,"composite":0.5431213026023877,"intelligence":7.598789169415934,"calibration":66.26655558235618,"speed":91.51497362122684,"cost_axis":80.09400261254245,"usd_per_1000_decisions":0.004608220443349753,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.05674442206509411,"p50_adjusted_s":0.26348884413018825,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"jul-fast","name":"jul fast (usejul/jul-decision-minicpm5-2b: MiniCPM5-2B fine-tune with a pointer head, option probabilities from one pass, self-hosted via `jul serve`)","board":"open","model_pin":"usejul/jul-decision-minicpm5-2b@f556a6936089693ea68bfed3080c637e4853a3ff","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/usejul/jul-decision-minicpm5-2b","listing":"ranked","ranked":true,"capability":42.76419280473996,"capability_eligible":true,"open_capability_rank":69,"composite":0.5174168972906218,"intelligence":7.4645122320968404,"calibration":78.06387337738309,"speed":90.69403874154062,"cost_axis":68.9870979950853,"usd_per_1000_decisions":0.010808449559918753,"price_kind":"estimate","price_basis":"ESTIMATE (size-class reference; self-hosted open weights, no bookable listing for this base). openbmb/MiniCPM5-2B is an EXPLICIT entry in the frozen module's UNLISTED_BASE_MODELS table, so pricing_v15.reference() returns None without raising: the decision that there is no bookable listing for this base is already taken, and the only MiniCPM key in either dated snapshot is deepinfra:openbmb/MiniCPM-Llama3-V-2_5 at (0.34, 0.34), a VISION model and not a comparable reference. The row therefore carries its own labelled size-class estimate: deepinfra:Qwen/Qwen3.5-2B at USD 0.02 per 1M input and USD 0.10 per 1M output in the frozen 25 Sep snapshot - a 2B dense instruct model, the same size class as this 2B base - x this row's own measured tokens. Estimated, not charged. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: the readout is a distribution over candidate labels from one forward pass and nothing is generated, so the row's cost rests on the input rate alone and any other decision is reversible by arithmetic from the raw. RELEASE-LANE LINE, recorded and not acted on: if a bookable MiniCPM5-2B text listing has appeared since 25 Sep, this row reprices by one multiplication from the raw. receipts/COST-ALTERNATIVES-R58.json has the arithmetic for the alternatives already done.","p50_raw_s":0.043561759011936374,"p50_adjusted_s":0.23712351802387274,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"opendecision","name":"OpenDecision (ModernBERT-large zero-shot)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/deepanwadhwa/OpenDecision","listing":"ranked","ranked":true,"capability":37.99564488451158,"capability_eligible":true,"open_capability_rank":75,"composite":0.508795165212109,"intelligence":7.425225288997898,"calibration":68.56606448002526,"speed":87.12856375997052,"cost_axis":79.10704862374335,"usd_per_1000_decisions":0.004970862068965518,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.07184981391765177,"p50_adjusted_s":0.29369962783530357,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"classone-gemma4-e2b","name":"ClassOne Gemma 4 E2B (Gemma-4-E2B backbone with ClassOne decision heads, schema-constrained single forward pass, self-hosted)","board":"open","model_pin":"devops-thiago/classone-gemma4-e2b@cd502aa33d288e67f16937c2e125f2297ab7b4ad","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/devops-thiago/classone-gemma4-e2b","listing":"ranked","ranked":true,"capability":15.208879168420996,"capability_eligible":true,"open_capability_rank":90,"composite":0.49351800738768026,"intelligence":7.802105373202749,"calibration":22.615652963639242,"speed":93.2878888255429,"cost_axis":70.20313863040258,"usd_per_1000_decisions":0.00984529451591063,"price_kind":"estimate","price_basis":"ESTIMATE (same-family reference, one size up; self-hosted open weights, no bookable listing for this base). pricing_v15.reference() RAISES on google/gemma-4-E2B and on google/gemma-4-e2b - the base is in none of the three frozen tables - and the published page carries no priced row on this base either: the one Gemma-4-E2B entry on the board, system-one-open, is listed not_measured. The row therefore carries its own labelled estimate from the nearest reference the frozen 25 Sep snapshot itself contains, deepinfra:google/gemma-4-E4B-it at USD 0.02 per 1M input and USD 0.10 per 1M output - the SAME Gemma-4 effective-parameter family, one size LARGER, so it is both the nearest available reference and the direction that cannot flatter this row - x this row's own measured tokens. Estimated, not charged. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, not by estimate: the readout is a distribution over candidate labels from one forward pass and nothing is generated, so the row's cost rests on the input rate alone and any other decision is reversible by arithmetic from the raw. RELEASE-LANE LINE, recorded and not acted on: google/gemma-4-E2B needs either a listing decision or an UNLISTED_BASE_MODELS entry in the frozen table before this row may be published; receipts/COST-ALTERNATIVES-R58.json has the alternatives' arithmetic already done.","p50_raw_s":0.029340580993448384,"p50_adjusted_s":0.20868116198689676,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"mghafiri-qwen3.5-0.8b-decision-model","name":"Qwen3.5-0.8B Decision Model (Mourad Ghafiri)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/mghafiri/qwen3.5-0.8B-decision-model","listing":"ranked","ranked":true,"capability":42.62692401157305,"capability_eligible":false,"open_capability_rank":null,"composite":0.47247471382875306,"intelligence":7.297270616183145,"calibration":77.95657740696296,"speed":55.90747410573378,"cost_axis":79.5180630553356,"usd_per_1000_decisions":0.00481649630541872,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate","p50_raw_s":5.551045611966401,"p50_adjusted_s":11.252091223932803,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"fastino-gliner-2-5-decide","name":"Fastino GLiNER-2.5-Decide (hosted API)","board":"api","model_pin":"fastino/GLiNER-2.5-Decide (requested model id)","last_measured_on":"2026-10-05","source_url":"https://fastino.ai/","listing":"ranked","ranked":true,"capability":34.57211625543646,"capability_eligible":true,"open_capability_rank":null,"composite":0.34332996901846624,"intelligence":7.011351364482595,"calibration":62.13288114639033,"speed":84.7052606179839,"cost_axis":45.81558630529788,"usd_per_1000_decisions":0.06399515910629654,"price_kind":"mixed","price_basis":"ESTIMATE: published Fastino sources conflict (USD 0.03/M vs 0.15/M input); conservative higher USD 0.15/M input, 0/M output x measured tokens; no invoice","p50_raw_s":0.4409300498664379,"p50_adjusted_s":0.4409300498664379,"latency_adjustment":"none (API)"},{"key":"deem-0.8-v1","name":"Deem 0.8B v1","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/LibertAIDAI/deem-0.8-v1","listing":"ranked","ranked":true,"capability":18.07951706793502,"capability_eligible":true,"open_capability_rank":88,"composite":0.32342918922131697,"intelligence":6.531204174677637,"calibration":29.6278299611924,"speed":85.44505463939521,"cost_axis":80.28784730320886,"usd_per_1000_decisions":0.004540166256157636,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: documented hosted-model estimate. Frozen market reference deepinfra:Qwen/Qwen3.5-0.8B.","p50_raw_s":0.15738983592018485,"p50_adjusted_s":0.4647796718403697,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"decision-fast","name":"Decision Fast (FlyMy.AI, v53a)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://huggingface.co/flymy-ai/decision-fast-preview","listing":"ranked","ranked":true,"capability":40.59403173266273,"capability_eligible":true,"open_capability_rank":72,"composite":0.2981489585308357,"intelligence":6.110691078442272,"calibration":75.07737238688318,"speed":91.472185793536,"cost_axis":80.08651867745036,"usd_per_1000_decisions":0.004610868226600985,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.05768406600691378,"p50_adjusted_s":0.2653681320138276,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"quyet-1-0-small-en","name":"Quyet-1.0-Small-EN (Chinh Nguyen, ModernBERT-base encoder with fixed heads)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/chinhnc/Quyet-1.0-Small-EN","listing":"ranked","ranked":true,"capability":39.53807694084099,"capability_eligible":true,"open_capability_rank":74,"composite":0.29239342204088814,"intelligence":6.054813340422714,"calibration":73.02134054125926,"speed":94.90893356488051,"cost_axis":89.11296503596958,"usd_per_1000_decisions":0.0023062222906403943,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): DeepInfra base-size encoders (bge-base, e5-base, gte-base, all-mpnet-base) $0.005/M input (ModernBERT-base has no market listing; base-size encoder reference) x the package's own measured usage.input_tokens; one forward pass, nothing generated; run in process on 1x H100 80GB (Lium), quyet 1.0.0; estimated, not charged","p50_raw_s":0.0145730719959829,"p50_adjusted_s":0.1791461439919658,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"laya-typed-decisions","name":"Laya typed-decisions","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/convaiinnovations/laya/tree/1c5edc17a7acd8701df6fc341c0d179f1c62c982/typed-decisions","listing":"ranked","ranked":true,"capability":44.22061775531203,"capability_eligible":false,"open_capability_rank":null,"composite":0.21666109526429145,"intelligence":5.471270906587037,"calibration":82.96996460403702,"speed":70.81965680870755,"cost_axis":82.52942826196218,"usd_per_1000_decisions":0.003822543103448276,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: documented hosted-model estimate; no exact base-model floor applies. $0.01/M input, $0/M output; same-class hosted-encoder estimate; no exact-base market floor.","p50_raw_s":0.8638670064901817,"p50_adjusted_s":1.8777340129803632,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"watersheep","name":"WaterSheep (Samrat Dutta, ModernBERT-base encoder with calibrated heads, served by the author's own `watersheep --serve`)","board":"open","model_pin":"samratduttaofficial/WaterSheep@ebf3705d2d9dc648d5cc6b4cbca7646c6ab48998","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/samratduttaofficial/WaterSheep","listing":"ranked","ranked":true,"capability":35.57380209564805,"capability_eligible":true,"open_capability_rank":78,"composite":0.19483758736149703,"intelligence":5.305239654184986,"calibration":65.84236453711111,"speed":77.5933742155104,"cost_axis":68.67136467637603,"usd_per_1000_decisions":0.011073574813811781,"price_kind":"estimate","price_basis":"ESTIMATE, labelled, and RUN 48'S RATE FOR THIS EXACT ROW AND BASE, CARRIED UNCHANGED: answerdotai/ModernBERT-base IS in the frozen pricing_v15.UNLISTED_BASE_MODELS, so no exact base-model floor applies and the module takes the estimate rate from the row's own registry entry; USD 0.03/M input, 0 output is the rate run 47 (quyet-10-small-en) and run 48 (watersheep on v1.5) already assigned to this base. Output 0 is literal: watersheep/infer.py returns output_tokens 0 and the serving path has no generation step (02 Oct deputy rule). x this row's own measured input tokens. Estimated, not charged. No new decision.","p50_raw_s":0.4014722369611263,"p50_adjusted_s":0.9529444739222527,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"openjev-verdict-1.4","name":"openJev Verdict 1.4","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/heman10x/rlcd-modernbert-151m","listing":"ranked","ranked":true,"capability":40.36740412437111,"capability_eligible":true,"open_capability_rank":73,"composite":0.17222636557331997,"intelligence":5.035901218168153,"calibration":75.69890703057408,"speed":81.49274954261848,"cost_axis":86.62313527343974,"usd_per_1000_decisions":0.002791871921182266,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: v1.5 measured-input proxy; no exact base-model floor applies","p50_raw_s":0.2926154159940779,"p50_adjusted_s":0.7352308319881559,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"raw-qwen3-0.6b","name":"Raw Qwen3 0.6B direct logits","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Qwen/Qwen3-0.6B","listing":"ranked","ranked":true,"capability":8.247746904337834,"capability_eligible":true,"open_capability_rank":92,"composite":0.1673536172919218,"intelligence":5.554864880982027,"calibration":10.94062892769364,"speed":93.52703208725202,"cost_axis":77.58073177551512,"usd_per_1000_decisions":0.005588676108374384,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.02898204093798995,"p50_adjusted_s":0.2079640818759799,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"lev-350m","name":"lev-350m (Franck Verrot, LFM2.5-350M)","board":"open","model_pin":null,"last_measured_on":"2026-10-05","source_url":"https://github.com/franckverrot/lev","listing":"ranked","ranked":true,"capability":43.34590358011229,"capability_eligible":true,"open_capability_rank":66,"composite":0.15377137469809216,"intelligence":4.827562367090417,"calibration":81.86424479313416,"speed":93.98230372494703,"cost_axis":80.03797760430007,"usd_per_1000_decisions":0.00462807881773399,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.022326925001834752,"p50_adjusted_s":0.1946538500036695,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"raw-qwen3-1.7b","name":"Raw Qwen3 1.7B direct logits","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Qwen/Qwen3-1.7B","listing":"ranked","ranked":true,"capability":8.320306683538828,"capability_eligible":true,"open_capability_rank":91,"composite":0.13671939226506033,"intelligence":5.124286554530066,"calibration":11.516326812547591,"speed":93.23530725162705,"cost_axis":68.54983190559568,"usd_per_1000_decisions":0.011177352216748768,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.03198186121881008,"p50_adjusted_s":0.21396372243762016,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"kev-0.5b","name":"kev 0.5B","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/jaredpalmer/kev","listing":"ranked","ranked":true,"capability":32.90230956430239,"capability_eligible":true,"open_capability_rank":81,"composite":0.131588623265787,"intelligence":4.598433868646836,"calibration":61.20618525995795,"speed":92.39904218127765,"cost_axis":80.09398520305423,"usd_per_1000_decisions":0.004608226600985222,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.043983193347230554,"p50_adjusted_s":0.2379663866944611,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"openjev-verdict","name":"openJev Verdict (heman10x, ModernBERT-base 151M)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Heman10x-NGU/openJev-verdict-2.0","listing":"ranked","ranked":true,"capability":26.63220276955003,"capability_eligible":true,"open_capability_rank":85,"composite":0.09650709186186587,"intelligence":4.1496399816416485,"calibration":49.11476555745841,"speed":79.40046783928,"cost_axis":86.62313527343974,"usd_per_1000_decisions":0.002791871921182266,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: v1.5 measured-input proxy; no exact base-model floor applies","p50_raw_s":0.2692721150815487,"p50_adjusted_s":0.6885442301630974,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"quyet-1-0-small","name":"Quyet-1.0-Small (Chinh Nguyen, SEA-LION-ModernBERT-300M encoder with fixed heads)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/chinhnc/Quyet-1.0-Small","listing":"ranked","ranked":true,"capability":32.492363520877504,"capability_eligible":true,"open_capability_rank":82,"composite":0.08963716950381444,"intelligence":4.019045994532796,"calibration":60.96568104722222,"speed":94.91846454857637,"cost_axis":79.56261245432609,"usd_per_1000_decisions":0.004800055418719211,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): DeepInfra same-size encoders (bge-large, e5-large, Qwen3-Embedding-0.6B) $0.01/M input (SEA-LION-ModernBERT-300M has no market listing; the ModernBERT-large-class reference of the Laya/Certo/OpenDecision rows) x the package's own measured usage.input_tokens; one forward pass, nothing generated; run in process on 1x H100 80GB (Lium), quyet 1.0.0; estimated, not charged","p50_raw_s":0.01451935400837101,"p50_adjusted_s":0.17903870801674202,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"watt-flash-0.1","name":"watt-flash-0.1 (Zaitgeist Labs, 140M mmBERT-small encoder with an option-marker scorer, one forward pass per request, self-hosted)","board":"open","model_pin":"zaitlabs/watt-flash-0.1@d49a3a880b0201b325f2d2299d4c36ffc8735477","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/zaitlabs/watt-flash-0.1","listing":"ranked","ranked":true,"capability":41.569072545871116,"capability_eligible":true,"open_capability_rank":71,"composite":0.07461802002995101,"intelligence":3.759262959387639,"calibration":79.37888213235459,"speed":75.84832436427908,"cost_axis":89.00253166733508,"usd_per_1000_decisions":0.0023258530805687202,"price_kind":"mixed","price_basis":"ESTIMATE (size-class reference; self-hosted open weights, no public tariff). THE SAME CHOICE THE PUBLISHED quyet-1-0-tiny ROW AND RUN 59'S tacet-sonata ROW CARRY FOR THE IDENTICAL mmBERT-small BACKBONE, carried unchanged and not a new judgement: DeepInfra base-size encoders at USD 0.005 per 1M input, 0 output - deepinfra:thenlper/gte-base, deepinfra:BAAI/bge-base-en-v1.5, deepinfra:intfloat/e5-base-v2 and deepinfra:sentence-transformers/all-mpnet-base-v2 are all (0.005, 0.0) in the frozen 25 Sep snapshot - x this row's own measured input tokens. Estimated, not charged. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, verified in the author's source: watt_flash/serve.py:answer() formats a softmax over the option markers of one encoder pass and the package contains no generation path at all (the 02 Oct deputy rule for classifier read-outs). THE SERVER REPORTS NO usage BLOCK OF ANY KIND, so the input counts come EXACTLY from the author's own unmodified watt_flash.layout.Layout.pack_request - the very number their own WattFlash.decide_batch returns as len(p) and only serve.py drops - under PREREG-R61-AMENDMENT-1, cross-checked three ways (independent reassembly of their documented layout; identity of their packer's Overflow set with the server's own 422 set), with no cost number at all on any disagreement; the counts sit in cost_estimate and never in usage. RELEASE-LANE LINE, recorded and not acted on, AND IT IS THE CARRIED quyet/tacet LINE, NOT A NEW ONE: jhu-clsp/mmBERT-small still has no explicit market-listing decision (pricing_v15.reference() raises), so an UNLISTED_BASE_MODELS entry is the release lane's.","p50_raw_s":0.4322121739387512,"p50_adjusted_s":1.0144243478775024,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"quyet-1-0-tiny","name":"Quyet-1.0-Tiny (Chinh Nguyen, mmBERT-small encoder (16 layers kept) with fixed heads)","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/chinhnc/Quyet-1.0-Tiny","listing":"ranked","ranked":true,"capability":31.08052871088324,"capability_eligible":true,"open_capability_rank":83,"composite":0.07064766318734213,"intelligence":3.6964771541368573,"calibration":58.46458026762962,"speed":95.09610910731021,"cost_axis":88.43590631389968,"usd_per_1000_decisions":0.0024292364532019703,"price_kind":"estimate","price_basis":"v1.5.8 fast-lane delivery row (combined v1.5.8 preview, not yet published) cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE (v1.5-M2 base-model reference; self-hosted open weights, no public tariff): DeepInfra base-size encoders (bge-base, e5-base, gte-base, all-mpnet-base) $0.005/M input (mmBERT-small 16-layer, 183M; base-size class reference, errs high vs the small-encoder $0.004 class) x the package's own measured usage.input_tokens; one forward pass, nothing generated; run in process on 1x H100 80GB (Lium), quyet 1.0.0; estimated, not charged","p50_raw_s":0.012666817499848548,"p50_adjusted_s":0.1753336349996971,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-jev-deberta-v3-large","name":"open-jev-deberta-v3-large (local CPU)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/kotoba-lang/typed-decisions","listing":"ranked","ranked":true,"capability":35.39766066495959,"capability_eligible":false,"open_capability_rank":null,"composite":0.03235814777127837,"intelligence":2.8293414417857914,"calibration":67.96597988813339,"speed":67.62138193768119,"cost_axis":77.5922354035203,"usd_per_1000_decisions":0.005583743842364532,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: v1.5 measured-input proxy; no exact base-model floor applies","p50_raw_s":1.6890084934420884,"p50_adjusted_s":3.5280169868841766,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"verdict-small","name":"verdict-small (Manavarya09, multilingual-e5-small 118M)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Manavarya09/verdict","listing":"ranked","ranked":true,"capability":37.63085686322157,"capability_eligible":true,"open_capability_rank":76,"composite":0.01899384266854915,"intelligence":2.3417095945172277,"calibration":72.92000413192591,"speed":89.48731594094681,"cost_axis":100,"usd_per_1000_decisions":0.000866152709359606,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.04910206887871027,"p50_adjusted_s":0.24820413775742053,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"tacet-sonata","name":"Tacet Sonata (CodePawl, 144M packed encoder on mmBERT-small, option-marker softmax in one pass, in-process via the author's tacet package)","board":"open","model_pin":"codepawl/tacet-sonata@26f447c35881a80020fe557c95ceb8c61bc3fd12","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/codepawl/tacet-sonata","listing":"ranked","ranked":true,"capability":24.310999978028352,"capability_eligible":true,"open_capability_rank":86,"composite":0.016405006835540154,"intelligence":2.2436273530567146,"calibration":46.37837260299999,"speed":80.28211114890141,"cost_axis":89.01467380421724,"usd_per_1000_decisions":0.0023236865267433986,"price_kind":"estimate","price_basis":"ESTIMATE (size-class reference; self-hosted open weights, no public tariff). THE SAME CHOICE THE PUBLISHED quyet-1-0-tiny ROW CARRIES FOR THE IDENTICAL mmBERT-small BACKBONE: DeepInfra base-size encoders at USD 0.005 per 1M input, 0 output - deepinfra:thenlper/gte-base, deepinfra:BAAI/bge-base-en-v1.5, deepinfra:intfloat/e5-base-v2 and deepinfra:sentence-transformers/all-mpnet-base-v2 are all (0.005, 0.0) in the frozen 25 Sep snapshot - x this row's own measured usage.input_tokens (the package's packed length, capped at 4,096). Estimated, not charged. OUTPUT TOKENS ARE 0 BY CONSTRUCTION, verified in the package source (tacet/decoding.py usage_of sets output_tokens 0; the readout is a softmax over option [MASK] markers from one encoder pass and nothing is generated), the 02 Oct deputy rule for classifier read-outs. RELEASE-LANE LINE, recorded and not acted on, AND IT IS THE QUYET ROWS' CARRIED LINE, NOT A NEW ONE: jhu-clsp/mmBERT-small still has no explicit market-listing decision (pricing_v15.reference() raises), so an UNLISTED_BASE_MODELS entry is the release lane's.","p50_raw_s":0.23136236518621445,"p50_adjusted_s":0.6127247303724289,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"laya","name":"Laya (Convai Innovations, ModernBERT-large 421M)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/convaiinnovations/laya","listing":"ranked","ranked":true,"capability":32.50714989618597,"capability_eligible":false,"open_capability_rank":null,"composite":0.008560099345041717,"intelligence":1.7910341507793501,"calibration":63.22326564159259,"speed":73.26302890816496,"cost_axis":84.89402784895724,"usd_per_1000_decisions":0.0031881034482758625,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.825308452360332,"p50_adjusted_s":1.800616904720664,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"mxbai-rerank-base-v2","name":"Mixedbread mxbai-rerank-base-v2","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/mixedbread-ai/mxbai-rerank-base-v2","listing":"ranked","ranked":true,"capability":43.87421595931999,"capability_eligible":true,"open_capability_rank":65,"composite":0.002273929358844764,"intelligence":1.140707327109041,"calibration":86.60772459153094,"speed":92.58758170313658,"cost_axis":60.33589314238483,"usd_per_1000_decisions":0.020996016009852216,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: size-class proxy (deepinfra:Qwen/Qwen3-Reranker-0.6B); no exact base-model floor applies","p50_raw_s":0.031425865134224296,"p50_adjusted_s":0.21285173026844859,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"laya-multilingual","name":"Laya multilingual","board":"open","model_pin":null,"last_measured_on":"2026-10-04","source_url":"https://huggingface.co/convaiinnovations/laya/tree/1c5edc17a7acd8701df6fc341c0d179f1c62c982/multilingual","listing":"ranked","ranked":true,"capability":16.971882566936873,"capability_eligible":true,"open_capability_rank":89,"composite":0.000037595668430455566,"intelligence":0.2879184668737643,"calibration":33.655846666999985,"speed":77.82044954853706,"cost_axis":82.1551651086044,"usd_per_1000_decisions":0.003933940886699508,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: documented hosted-model estimate; no exact base-model floor applies. $0.01/M input, $0/M output; same-class hosted-encoder estimate; no exact-base market floor.","p50_raw_s":0.3577279045130126,"p50_adjusted_s":0.8654558090260253,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"certo","name":"Certo v1 (AltSlate Labs)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/altslate/certo-decision-model","listing":"ranked","ranked":true,"capability":43.17437163074946,"capability_eligible":true,"open_capability_rank":67,"composite":0.000013254531668229452,"intelligence":0.20278658812856143,"calibration":86.14595667337036,"speed":92.53493469610669,"cost_axis":96.9130801359906,"usd_per_1000_decisions":0.0012673522167487683,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.042036901926621795,"p50_adjusted_s":0.23407380385324358,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"clm-8b","name":"CLM-8B (Contrastive-LM, clm-latest)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Contrastive-LM/CLM","listing":"ranked","ranked":true,"capability":20.086328406467768,"capability_eligible":true,"open_capability_rank":87,"composite":0.0000035510969933114796,"intelligence":0.13075619159747265,"calibration":40.041900621338065,"speed":92.63548383949004,"cost_axis":50.501662240594015,"usd_per_1000_decisions":0.044662660714285714,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): price floor: base-model reference price applied","p50_raw_s":0.02119073923677206,"p50_adjusted_s":0.19238147847354412,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"bge-reranker-v2-m3","name":"BAAI bge-reranker-v2-m3","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/BAAI/bge-reranker-v2-m3","listing":"ranked","ranked":true,"capability":44.08279577608597,"capability_eligible":true,"open_capability_rank":63,"composite":4.4076803151681546e-7,"intelligence":0.06512222845080358,"calibration":88.10046932372114,"speed":93.86976236118387,"cost_axis":59.325605115070715,"usd_per_1000_decisions":0.022688885467980296,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: base-model market reference (deepinfra:BAAI/bge-m3)","p50_raw_s":0.01707328436896205,"p50_adjusted_s":0.1841465687379241,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"gte-reranker-modernbert-base","name":"Alibaba GTE Reranker ModernBERT-base","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://huggingface.co/Alibaba-NLP/gte-reranker-modernbert-base","listing":"ranked","ranked":true,"capability":41.7278116825546,"capability_eligible":true,"open_capability_rank":70,"composite":0,"intelligence":0,"calibration":83.4556233651092,"speed":93.76120791185616,"cost_axis":69.26965431492843,"usd_per_1000_decisions":0.010576570197044334,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: size-class proxy (deepinfra:thenlper/gte-base); no exact base-model floor applies","p50_raw_s":0.024179050931707025,"p50_adjusted_s":0.19835810186341404,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"mirror","name":"Mirror","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/Bluusun/Decision-API","listing":"ranked","ranked":true,"capability":25.806386196486883,"capability_eligible":false,"open_capability_rank":null,"composite":0,"intelligence":0,"calibration":51.61277239297377,"speed":73.47211485993975,"cost_axis":89.27300805786774,"usd_per_1000_decisions":0.0022780665024630543,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.6508294129744172,"p50_adjusted_s":1.4516588259488343,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"open-jev-json-canvas-joshuasp","name":"Open Jev JSON Canvas (JoshuaSP)","board":"open","model_pin":null,"last_measured_on":"2026-10-02","source_url":"https://github.com/JoshuaSP/open-jev","listing":"ranked","ranked":true,"capability":30.596109035139946,"capability_eligible":true,"open_capability_rank":84,"composite":0,"intelligence":61.19221807027989,"calibration":0,"speed":85.53332924416175,"cost_axis":49.295219536461964,"usd_per_1000_decisions":0.04899585591133005,"price_kind":"estimate","price_basis":"v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): documented hosted-model estimate; no exact base-model floor applies","p50_raw_s":0.14833511249162257,"p50_adjusted_s":0.44667022498324516,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"apus-openjev-v1-35b-a3b","name":"APUS-OpenJev-v1-35B-A3B (merged bf16 checkpoint-5949 MoE on Qwen3.5-35B-A3B, their own native runtime, full depth, self-hosted)","board":"open","model_pin":"apus-ailab/APUS-OpenJev-v1-35B-A3B@090388abe2237117cc3b4df6f4250e81cbb36864","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/apus-ailab/APUS-OpenJev-v1-35B-A3B","listing":"listed","ranked":false,"capability":60.36229840333269,"capability_eligible":false,"open_capability_rank":null,"composite":null,"intelligence":52.808647095759525,"calibration":67.91594971090585,"speed":80.50610262660216,"cost_axis":null,"usd_per_1000_decisions":null,"price_kind":"unpriced","price_basis":"UNPRICED, and deliberately not invented. pricing_v15.reference() RAISES on the exact declared base Qwen/Qwen3.5-35B-A3B ('base model lacks an explicit market-listing decision'). Run 46's bongard-mini rule applies unchanged: the frozen module raises rather than estimating, and the FROZEN-TABLE ROW is the release lane's (board #11 / #10132) - not this job's and not Florian's, because his standing 24 Sep 20:30 rule already decides the principle and the 26 Sep rule fixes the method. Complete but unpriced - shown, not ranked. RELEASE-LANE LINE, and a tight one: the dated snapshot ALREADY lists the exact base as qwen/qwen3.5-35b-a3b at (0.3125, 1.25), and BASE_REFERENCES ALREADY carries 'Qwen/Qwen3.5-35B-A3B-Base' -> 'qwen/qwen3.5-35b-a3b'. Only the non--Base (instruct) spelling of the same model is missing, so every Qwen3.5-35B-A3B derivative raises today. BECAUSE OUTPUT TOKENS ARE 0 BY CONSTRUCTION this row's Cost depends on the input rate alone, so the number is one multiplication over tokens the raw keeps; r56/receipts/COST-ARITHMETIC-R56.json carries it already computed.","p50_raw_s":0.3547162465401925,"p50_adjusted_s":0.859432493080385,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"bb-qwen3.5-4b-lora","name":"BB-Qwen3.5-4B-LoRA (Babak Barazandeh, LoRA on Qwen3.5-4B-Base, hosted /v1/systemone)","board":"api","model_pin":"checkpoint-030500 (author declaration); base revision 1001bb4d","last_measured_on":"2026-10-07","source_url":"https://github.com/fstandhartinger/jevbench/issues/156","listing":"listed","ranked":false,"capability":60.40606706607417,"capability_eligible":false,"open_capability_rank":null,"composite":null,"intelligence":38.67355961154223,"calibration":82.13857452060611,"speed":81.57813729165589,"cost_axis":null,"usd_per_1000_decisions":null,"price_kind":"estimated","price_basis":"UNPRICED - NOT A MEASUREMENT FAILURE. The base is disclosed (Qwen/Qwen3.5-4B-Base; reference() resolves it to deepinfra:Qwen/Qwen3.5-4B at 0.03/0.15), but the endpoint reports usage.input_tokens ONLY - no output_tokens on any answer - and the system is API-only, so 'nothing is generated' cannot be verified in its code; the 02 Oct deputy rule keeps an unverified null unpriced. The frozen scorer therefore gives no Cost axis and no composite; the row is reported on Capability. RELEASE-LANE LINE: if the author confirms (or the release lane accepts) a one-pass readout with 0 output tokens, the row prices by one multiplication from the raw at 0.03/M input (receipts/COST-ALTERNATIVES-R59.json).","p50_raw_s":0.6954362019896507,"p50_adjusted_s":0.6954362019896507,"latency_adjustment":"none (API)"},{"key":"seb-9b","name":"Seb-9B (Qwen3.5-9B-shaped multimodal decision model, one forward pass, stock vLLM, self-hosted)","board":"open","model_pin":"ironbcc/seb-9b@ac57b36bbe978762032d0ad7197778b4bc4a0209","last_measured_on":"2026-10-07","source_url":"https://huggingface.co/ironbcc/seb-9b","listing":"listed","ranked":false,"capability":66.79019043230528,"capability_eligible":false,"open_capability_rank":null,"composite":null,"intelligence":43.512227190846936,"calibration":90.06815367376362,"speed":90.02719232031932,"cost_axis":null,"usd_per_1000_decisions":null,"price_kind":"unpriced","price_basis":"UNPRICED, and deliberately not invented. The author declares NO base model anywhere in the repository, so pricing_v15.reference() has nothing to resolve: it returns None without raising, and the frozen module records the row unpriced. Complete but unpriced - shown, not ranked - exactly as run 52 left ines-1. RELEASE-LANE LINE: whether an UNDECLARED base that is architecturally identified may supply a size-class estimate. The repo ships Qwen3_5ForConditionalGeneration with 32 layers, hidden 4096 and the Qwen3.5 tokenizer under the name '9B', and qwen/qwen3.5-9b = (0.10, 0.15) is the hosted model of that exact shape; r56/receipts/COST-ARITHMETIC-R56.json carries the resulting cost per 1,000 over this row's own measured tokens, clearly labelled as OUR candidate reference and NOT the author's claim, so the release lane needs no re-measurement either way.","p50_raw_s":0.06013961852295324,"p50_adjusted_s":0.2702792370459065,"latency_adjustment":"x2 + 0.15 s (assumption, not measured)"},{"key":"wity-1-always","name":"wity-1 (Wity, reasoning always)","board":"api","model_pin":null,"last_measured_on":"2026-10-06","source_url":"https://wity.alphanimble.com/","listing":"listed","ranked":false,"capability":79.25057518952474,"capability_eligible":false,"open_capability_rank":null,"composite":70.56449594359258,"intelligence":69.25112551166924,"calibration":89.25002486738025,"speed":71.20296182763504,"cost_axis":58.8349720563948,"usd_per_1000_decisions":0.023559582938388626,"price_kind":"tariff","price_basis":"Wity's stated API tariff USD 0.042 per 1M input tokens, output free; measured on the v1.6.1 common cost basis (all items except those outside the Jev reference input range) from this run's token usage","p50_raw_s":1.6885605081915855,"p50_adjusted_s":1.6885605081915855,"latency_adjustment":"none (API)"},{"key":"wity-1-off","name":"wity-1 (Wity, reasoning off)","board":"api","model_pin":null,"last_measured_on":"2026-10-06","source_url":"https://wity.alphanimble.com/","listing":"listed","ranked":false,"capability":62.968426842803666,"capability_eligible":false,"open_capability_rank":null,"composite":49.37499462500363,"intelligence":44.11116118075945,"calibration":81.82569250484788,"speed":89.55912090143326,"cost_axis":58.8349720563948,"usd_per_1000_decisions":0.023559582938388626,"price_kind":"tariff","price_basis":"Wity's stated API tariff USD 0.042 per 1M input tokens, output free; measured on the v1.6.1 common cost basis (all items except those outside the Jev reference input range) from this run's token usage","p50_raw_s":0.28173011541366577,"p50_adjusted_s":0.28173011541366577,"latency_adjustment":"none (API)"}]}