imajev 9B
adapter · 9BBase model: Qwen/Qwen3.5-9Bsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
34.524
Rank #3 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 48.3
- calibration
- 85.8
- speed
- 80.7
- cost
- 40.1
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.088717
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.259 s / 0.558 s
- Adjusted latency · p50 / p95
- 0.669 s / 1.265 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 59.6%
- Sealed accuracy
- 68.2%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.