Step 3.7 Flash: benchmark results
StepFun's speed-oriented Step 3.7 Flash model. Provider: StepFun. Released 2026-05-29. Access: Open.
Unified ELO 1653 ± 1, rank #126 of 1392 rated models, from 81 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA TAU-2 Bench | 98.54 | Accuracy (%) | 99 |
| AA Omniscience - Software Engineering (SWE) - Rust | 70 | Accuracy (%) | 92.2 |
| CLAW-Eval | 80.5 | Score (%) | 92 |
| OpenRouter Tau2-Bench Airline | 77.3 | Accuracy (%) | 92 |
| OpenClawProBench | 72 | Overall Score (%) | 90.9 |
| AA Omniscience - Software Engineering (SWE) - Julia | 32 | Accuracy (%) | 90.1 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 34 | Accuracy (%) | 88.1 |
| AA Omniscience - Software Engineering (SWE) - HTML | 50 | Accuracy (%) | 87.7 |
| AA Terminal-Bench Hard | 35.61 | Accuracy (%) | 82 |
| AA Omniscience - Software Engineering (SWE) - Python | 30.5 | Accuracy (%) | 80.9 |
| AA IFBench | 67.28 | Accuracy (%) | 80.4 |
| AA Omniscience - Software Engineering (SWE) - Java | 21 | Accuracy (%) | 77.3 |
Interactive version: theaggregate.ai/model?slug=step-3-7-flash · How It Works · Data refreshed daily, snapshot 2026-09-05.