jev-spatial (Fr0zencr4nE)
adapterBase model: allenai/Molmo2-ERsourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
1.970
Rank #55 of 65 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 12.7
- calibration
- 39.7
- speed
- 89.7
- cost
- 58.8
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.023654
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.068 s / 0.110 s
- Adjusted latency · p50 / p95
- 0.286 s / 0.370 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 46.8%
- Sealed accuracy
- 40.8%
Published 2026-11-17. Scores and ranks can change in a later release. See the method notes.