Step 3.7 Flash
| Released | 2026-05-29 |
|---|---|
| Price per 1M | $0.20 in · $1.21 out |
| Overall score | 49.1 · rank 120 |
| Context window | 256,000 tokens |
| Provider | StepFun |
| Weights | Open weights |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 49.1 | 120 | 5 |
| Overall: Coding | 46.9 | 134 | 3 |
| Overall: Reasoning | 51.5 | 101 | 3 |
| Overall: Agentic work | 47.8 | 132 | 3 |
| Overall: Frontend | 46.2 | 125 | 1 |
| Coding | 46.2 | 145 | 1 |
| Reasoning | 50.4 | 112 | 2 |
| Math | 52.8 | 90 | 1 |
| General | 50.9 | 95 | 1 |
| Value vs price | +3.8 | 89 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 48.5 ±4.3 | 3 |
| High | 50.9 ±5.4 | 2 |
Evidence
Coding 46.2 ±4.7
Reasoning 50.4 ±3.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Extended NYT Connections | 39.7% | Lech Mazur | 2026-11-10 |
| CritPt | 2.3% | Epoch AI | 2026-11-11 |
Math 52.8 ±4.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| MathArena Final-Answer Competitions | 68.5% | MathArena | 2026-11-11 |
General 50.9 ±3.3
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LLM Debate Benchmark | 1,374 | Lech Mazur | 2026-10-26 |
Names matched to this model
step-3-7-flash