Wity-1 (AlphaNimble), reasoning off
Closed APIBase model: undisclosedthe operator's hosted API
Image JevBench v0.3.0 composite score
33.085
Rank #6 of 52 ranked systems.
From the published full-benchmark aggregate. The composite combines intelligence, calibration, speed and cost.
Published axes
- intelligence
- 37.4
- calibration
- 74.9
- speed
- 72.5
- cost
- 71.8
Cost and speed evidence
- Cost per 1,000 decisions
- USD 0.008715
- Cost basis
- OpenRouter usage.cost receipts
- Measured latency · p50 / p95
- 1.639 s / 3.443 s
- Adjusted latency · p50 / p95
- 1.639 s / 3.443 s
- Latency adjustment
- none (hosted API, measured as is)
- Public accuracy
- 57.0%
- Sealed accuracy
- 59.8%
- Measurement note
- Official ImageJev v0.3 A300+P300 pass 1 on 7 Oct 2026 through the live Wity SystemOne endpoint (build 47ac68ed844f on all 600 responses), second Wity-1 mode row requested by AlphaNimble; same request as the reasoning auto row except reasoning=off. 600/600 ok, no retries.
Published 2026-10-07. Scores and ranks can change in a later release. See the method notes.