Image JevBench v0.1.5 · individual system
NeoHorse Jev 4B
4BBase model: NeoHorse-1-4BsourceArchived v0.1.5 score, not ranked on the v0.3 scale. See the v0.1.5 archive.
local/GPU evaluation
Image JevBench v0.1.5 composite score
71.937
v0.1.5 rank #4 of 50 (archived).
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 72.9
- calibration
- 91.2
- speed
- 86.8
- cost
- 51.6
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.041177
- Cost basis
- measured GPU seconds x $1.19/GPU-hour
- Measured latency · p50 / p95
- 0.140 s / 0.170 s
- Adjusted latency · p50 / p95
- 0.431 s / 0.490 s
- Latency adjustment
- v1.4 local latency adjustment
- Public accuracy
- 80.7%
- Sealed accuracy
- 82.7%
Published 2026-10-01. Scores and ranks can change in a later release. See the method notes.