Model · Stepfun
Stepfun Step Flash 3.7
2 appearances · meta-score 66.5/100 · last seen 2026-07-19 · scored under rubric v15
Pooled composite across 2 primary runs: 57.9 ± 12.7 (95% CI) provisional: n < 3, treat as unsettled
- $70.42Total cost
- 1,011,239,246Tokens
- 87.3%Cache hit
- 16,120Requests
- 0.07%Failure rate
- 1.12Tools/response
- 14,360,879Tokens per $
Agent role scores
| Agent role | Score / 10 |
|---|---|
| Architect | 7.80 |
| Orchestrator | 7.53 |
| Scout Agent | 7.33 |
| Build Agent | 7.26 |
| Sprint-Review Agent | 7.21 |
| Sprint-Plan Agent | 7.17 |
| Develop Agent | 7.10 |
| Test Agent | 7.08 |
| Platform Agent | 6.80 |
| Pseudocode Agent | 6.71 |
| Code Agent | 6.12 |
| Code-Review Agent | 5.89 |
| Director | 4.36 |
Run history (primary model)
| Run | SoW | Date | Composite |
|---|---|---|---|
| Step Flash 3.7 | circles | Jul 2026 | 58.9 |
| Step Flash 3.7 | solarSystem | Jul 2026 | 56.9 |
Every run on Favur Evals is scored by the same deterministic engine, on the same Statements of Work. The benchmark is self-funded — no vendor sponsorship, credits, or grants.