Image JevBench v0.3.0 · individual system
JPT-9B (kirp / llm2jev)
adapter · 9BBase model: Qwen3.5-9Bsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
31.435
Rank #5 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 44.3
- calibration
- 73.2
- speed
- 81.6
- cost
- 42.5
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.082697
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.225 s / 0.501 s
- Adjusted latency · p50 / p95
- 0.601 s / 1.152 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 61.7%
- Sealed accuracy
- 62.8%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.