Image JevBench v0.3.0 · individual system
PlayJev 0.8B
fine-tune · 0.8BBase model: Qwen/Qwen3.5-0.8B-Basesourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
0.502
Rank #39 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 7.6
- calibration
- 43.6
- speed
- 87.3
- cost
- 53.9
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.034425
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.092 s / 0.206 s
- Adjusted latency · p50 / p95
- 0.335 s / 0.562 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 38.7%
- Sealed accuracy
- 37.2%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.