Image JevBench v0.3.0 · individual system
Bonsai-2-27B v2 PQ2_0 + Q8_0 MMProj
LLM decoderBase model: undisclosedsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
11.983
Rank #21 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 43.2
- calibration
- 82.9
- speed
- 72.9
- cost
- 28.9
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.233998
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.652 s / 1.701 s
- Adjusted latency · p50 / p95
- 1.453 s / 3.552 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 63.8%
- Sealed accuracy
- 60.6%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.