Torchcast Decision 31B v1.0v1.7 · 10 s rule
Data: JevBench v1.7.14, published 2026-10-28 · method v1.7 (10 s rule), measured under the rule
Capability 91.8 · Not ranked in the current release
Measured on the 27 Oct 2026 draw, which is not yet equated to the reference scale. This draw differs strongly from earlier draws (72% multilingual items). Shown unranked until the draw is equated.
Last measured: 2026-10-27 · measurement release: v1.7.14
Apache-2.0 (weights and serving package; base Gemma-4-31B-it Apache-2.0), public, not gated
intelligence
88.4
calibration
95.3
speed
87.3
cost
52.4
Composite (secondary)
76.4 · Not ranked on this board
Cost and measurement conditions
$0.039 per 1,000 decisions (estimated)
ESTIMATE (*est.), self-hosted open weights, no public tariff: base google/gemma-4-31B-it, google/gemma-4-31b-it at USD 0.09/0.34 per million input/output tokens (frozen pricing module sha256 645fd82eac728d597048bec0995ecd3ac896566e269fb6c9a1fdb84c171a9ac4). Input is billed on the engine-processed prompt tokens as reported by the server (equal to our offline count on every item): 645,268 tokens on 1,500 answered items; the 5 head+tail-cut states count at their truncated size. Output: 0 tokens (one forward pass per question, option-letter readout, no generation). COST FIGURES: USD 0.038716 per 1,000 decisions over all 1,500 items of the draw (every row on this draw is costed on all 1,500 items); 0.60x the official cost cap USD 0.064595. Estimated model tariff, not charged usage or a host quote.
Measured p50 latency: 0.095 s (on-GPU, no network, loopback; not directly comparable with API latency, which we time end to end from our measuring host, network included)
Ranking and cap use: 0.34 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.095 s on-GPU, no network (timed on our GPU over loopback). Not directly comparable with API latency, which we time end to end from our measuring host, network included. The ranking and latency cap use the same estimate for every self-hosted row: 2 × measured + 150 ms = 0.34 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
Evaluator-owned rented Verda GPU VM, 1x RTX PRO 6000 Blackwell Server Edition 96 GB (1RTXPRO6000.30V, FIN-03, Finland, EU, pay-as-you-go non-spot, USD 2.608/h), instance id c92731d5-54bc-4f2e-b781-5434544f75f0. Container with --network none (cut proven from inside before the sealed input), Hub offline, every credential unset, read-only weights, all 14 weight shards sha256-verified; hash-locked torch 2.13.0+cu130 / transformers 5.17.0 stack; driven one request at a time. The standard self-hosted x2 + 0.15 s adjustment applies. Measured 27 Oct 2026 on the frozen v1.6-regular-20261027 self-hosted input (sha256 9b2f32ed...); determinism 50/50 bit-identical; VM and volume destroyed afterwards.
Measured latency over 1,500 decisions: p50 0.095 s · p95 0.202 s
0 decisions over 10 s.
A decision that takes longer than 10 seconds (measured time) counts as a wrong answer and as zero calibration for that item: each type's calibration is multiplied by the share of its decisions answered within 10 s. The limit uses the measured time (raw). The displayed adjusted time for self-hosted rows (2 × + 0.15 s) is an assumption and is not used for the limit.