Image JevBench v0.3.0 · individual system
Mapika decider-2b-vision BF16
fine-tune · 2BBase model: Qwen/Qwen3.5-2B-Basesourcelocal/GPU evaluation
Image JevBench v0.3.0 composite score
9.520
Rank #23 of 44 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 23.1
- calibration
- 62.6
- speed
- 85.8
- cost
- 54.0
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.034199
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.117 s / 0.271 s
- Adjusted latency · p50 / p95
- 0.384 s / 0.691 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 48.9%
- Sealed accuracy
- 47.7%
Published 2026-10-06. Scores and ranks can change in a later release. See the method notes.