Image JevBench v0.3.0 · individual system
Autoloops – Gemma 4 31B IT
Closed APIBase model: undisclosedthe operator's hosted API
Image JevBench v0.3.0 composite score
41.484
Rank #4 of 55 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 49.7
- calibration
- 87.1
- speed
- 75.0
- cost
- 42.4
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.083403
- Cost basis
- OpenRouter usage.cost receipts
- Measured latency · p50 / p95
- 1.185 s / 2.662 s
- Adjusted latency · p50 / p95
- 1.185 s / 2.662 s
- Latency adjustment
- none (hosted API, measured as is)
- Public accuracy
- 67.2%
- Sealed accuracy
- 67.1%
- Measurement note
- Official ImageJev v0.3 A300+P300 pass 1 on 7 Oct 2026 via api.autoloops.ai /v1/systemone (gemma-4-31b-it), request mapping of the hash-pinned 24 Sep Autoloops runner (the exact v0.2 clean runner file is missing, so byte parity with v0.2 is not claimed). 600/600 ok, no retries. Images sent with their real MIME type; BMP re-encoded losslessly as PNG; images over 6 MiB downscaled to 2,048 px.
Published 2026-10-12. Scores and ranks can change in a later release. See the method notes.