Image JevBench v0.3.0 · individual system
AutoJev-27B
fine-tune · 27BBase model: Qwen/Qwen3.8-27Bsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
29.339
Rank #7 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 37.6
- calibration
- 86.7
- speed
- 85.7
- cost
- 47.9
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.054405
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.167 s / 0.204 s
- Adjusted latency · p50 / p95
- 0.484 s / 0.558 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 60.4%
- Sealed accuracy
- 56.5%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.