Image JevBench v0.3.0 · individual system
Winnow-12B (Q8_0 + F16 vision projector)
merge · 12B · GGUF Q8_0Base model: google/gemma-4-12B-itsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
17.076
Rank #15 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 38.6
- calibration
- 61.4
- speed
- 82.5
- cost
- 38.0
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.086730
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.267 s / 0.333 s
- Adjusted latency · p50 / p95
- 0.684 s / 0.817 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 58.7%
- Sealed accuracy
- 58.6%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.