decisio v0.8.0 on gemma-4-12B-it (frozen, 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 70.7 · rank #78 of 251 on the open-weights board

Last measured: 2026-10-06 · measurement release: v1.6.1

Apache-2.0 software; Gemma model terms

Base model: google/gemma-4-12B-itsource

intelligence

52.3

calibration

89.2

speed

86.3

cost

67.7

Composite (secondary)

70.6 · Composite rank #77

Cost and measurement conditions

$0.012 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: 779,965 input + 1,477 output tokens over 1,477 decisions = 529.07379824 tokens/decision. GPU-time cost USD 0.0119075160195099/1,000; old list-price cost USD 0.0265536899119838/1,000; applied min(old list-price cost, GPU-time cost) = USD 0.0119075160195099/1,000. 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: 0.116 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: 0.38 s (est. incl. network (2x + 150 ms))

Self-hosted model: measured p50 0.116 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 = 0.38 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. github.com/aminry/decisio tag v0.8.0 (5b42101a191d222062a795194b0aedb5003bf6ea, verified against the GitHub refs API), installed with their own `uv sync --extra serve --frozen` so their lockfile fixes every version, and served with their own `python -m decisio.serve.vllm_engine --base gemma-4-12b`. --base carries google/gemma-4-12B-it@707f0a3b (bf16, 22.8 GiB) and that profile's measured settings - temperature 3.592 for every question type, a system turn, the chat template's own answer position, every single-token form of each letter summed. Frozen base: no fine-tuning, no adapter, no task registered. One RTX 6000 Ada 48 GB (Lium, Toronto). Latency axis uses the standard self-hosted x2 + 0.15 s adjustment. add-requests run 54, GitHub #169

Measured latency over 1,500 decisions: p50 0.116 s · p95 0.231 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 & law46.1n=1616
Coding & software71.0n=382
Math & numbers9.2n=370
Finance & commerce48.7n=223
Support & operations47.3n=160
Everyday language75.1n=145
Safety & security70.7n=109

Use cases (TypeSafe categories)

Model routing73.7n=524
Legal & compliance52.6n=515
Customer support47.5n=240
E-commerce48.9n=119
Insurance claims58.1n=119
Risk assessment33.0n=118
Financial crime21.5n=101
Feature extraction26.1n=99
Lead generation51.6n=95
Recruiting23.7n=94
Knowledge graphs47.3n=94
LLM guardrails64.6n=93
Moderation56.7n=93
Code linting52.3n=90
Search & retrieval78.1n=90
Science16.0n=90
Advertising41.6n=88
Gaming7.7n=86
Demand forecasting0.0n=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 70.74 · Composite 68.17 · rank #36 on the all-systems board · #17 open weights

intelligence 52.32 · calibration 89.16 · speed 86.99 · cost 59.35

v1.6 speed basis (124 items): p50 0.102 s · p95 0.207 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.

Published source