Image JevBench v0.3.0 · individual system
CUA-S1 4B
adapter · 4BBase model: Qwen/Qwen3.5-4Bsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
12.894
Rank #20 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 30.2
- calibration
- 68.9
- speed
- 81.9
- cost
- 42.9
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.080320
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.218 s / 0.481 s
- Adjusted latency · p50 / p95
- 0.587 s / 1.112 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 58.7%
- Sealed accuracy
- 52.2%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.