Hopper 12B trained (gemma-4-12B-it frozen + LoRA r32 unmerged, one forward pass, self-hosted)v1.7 · 10 s rule
Data: JevBench v1.8.18, published 2026-11-20 · method v1.7 (10 s rule), re-scored from the recorded per-decision times
Capability 68.1 · rank #97 of 251 on the open-weights board
Last measured: 2026-10-06 · measurement release: v1.6.1
apache-2.0
intelligence
51.7
calibration
84.6
speed
83.6
cost
66.6
Composite (secondary)
68.8 · Composite rank #90
Cost and measurement conditions
$0.013 per 1,000 decisions (estimated)
GPU-time estimate from v1.8.0: one monthly-rented RTX PRO 6000 96 GB, EUR 999/month / 730 hours, ECB USD 1.1133/EUR dated 28 Oct 2026, USD 1.5235/hour; average load 70%. Architecture: small gemma, 12B active parameters; throughput 26861.9166667 tokens/s (architecture-class calibration, 29 Oct 2026, efficient-engine prefill; capped at 100,000 tokens/s). The row's own scored cost basis: 854,727 input + 0 output tokens over 1,477 decisions = 578.69126608 tokens/decision. GPU-time cost USD 0.0130242237361284/1,000; old list-price cost USD 0.0289345633039946/1,000; applied min(old list-price cost, GPU-time cost) = USD 0.0130242237361284/1,000. The GPU-time estimate is lower. Estimated, not a bill. Intelligence, Calibration, measured latency and the common cost-item membership are unchanged. Prior prices remain in the archived v1.7 releases.
Measured p50 latency: 0.183 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: 0.52 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.183 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 = 0.52 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
measured 2026-10-06 on the live v1.6.0/v1.6.1 pool. Serving code github.com/hopit-ai/hopper tag 12b-1.1.1, installed with the author's own one-line pip install; adapter HopitAI/hopper@4b605f25eb903c2cf4d121040ffb097240b3bc37 (branch hopper-12b), LoRA r32 alpha 64, loaded UNMERGED because the author states a bf16 merge changes answers; base google/gemma-4-12B-it@707f0a3b, frozen; readout map maps/hopper-12b-trained-readout.json as shipped in their package. 1.1.1 is their own fix for corrupted option scores at prompt lengths one above a multiple of 32. Their flags left at their defaults: --shortlist tournament (their measured default) for menus over 26 options, --cuda-graphs OFF. The author asked on 6 Oct 2026 that the TRAINED version replace the untrained Hopper 12B request in GitHub #164; this is that row. NOT the published `hopper` row (HopitAI, rank 25), a different system. One RTX 6000 Ada 48 GB (Lium, Toronto); the author ran theirs on an L40S. Latency axis uses the standard self-hosted x2 + 0.15 s adjustment. add-requests run 54, GitHub #164
Measured latency over 1,500 decisions: p50 0.183 s · p95 0.344 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.
Breakdowns (not part of the score)
Chance-corrected competence (0 = chance, 100 = perfect) by category on S+P+L1+L2+L3. Raw values, not equated; the same cells feed the compare view.
Capability by subject topic
| Rules, policy & law | 46.1 | n=1616 |
|---|---|---|
| Coding & software | 64.8 | n=382 |
| Math & numbers | 7.8 | n=370 |
| Finance & commerce | 48.4 | n=223 |
| Support & operations | 47.1 | n=160 |
| Everyday language | 70.5 | n=145 |
| Safety & security | 62.1 | n=109 |
Use cases (TypeSafe categories)
| Model routing | 69.5 | n=524 |
|---|---|---|
| Legal & compliance | 54.3 | n=515 |
| Customer support | 42.7 | n=240 |
| E-commerce | 44.4 | n=119 |
| Insurance claims | 54.2 | n=119 |
| Risk assessment | 42.0 | n=118 |
| Financial crime | 17.8 | n=101 |
| Feature extraction | 22.6 | n=99 |
| Lead generation | 50.9 | n=95 |
| Recruiting | 30.7 | n=94 |
| Knowledge graphs | 45.6 | n=94 |
| LLM guardrails | 76.9 | n=93 |
| Moderation | 62.9 | n=93 |
| Code linting | 37.3 | n=90 |
| Search & retrieval | 62.0 | n=90 |
| Science | 14.0 | n=90 |
| Advertising | 23.9 | n=88 |
| Gaming | 19.3 | n=86 |
| Demand forecasting | 0.0 | n=83 |
Cells need at least 30 answered items; † fewer than 30 (indicative only; drawn as hollow, lighter points on the compare radars).
JevBench v1.6.43 values (method v1.6, before the 10 s rule)
Capability 68.15 · Composite 65.78 · rank #47 on the all-systems board · #26 open weights
intelligence 51.66 · calibration 84.63 · speed 84.57 · cost 56.16
v1.6 speed basis (124 items): p50 0.164 s · p95 0.290 s
These figures are kept for comparison. The current figures above are re-scored under the 10 s rule from the recorded per-decision times and replace them.