StepFunStep 3.5 Flash
Best configuration ranks #189 of 610 on the BenchLeader Index at 54.9 ±3.7. Last measured 2 Sept 2026. Released 11 Feb 2026.
- Blended price
- $0.150/M
- $0.100 in · $0.300 out
- Output speed
- 216 tok/s
- First answer
- 10 s
- first token 1.30 s
- Context
- 256k
- Overall index55
- Reasoning58
- Coding59
- Agents & tools62
- Maths52
- Knowledge44
- Instruction following65
- Human preference58
- Long context58
- Composite50
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | default | Source |
|---|---|---|
| LMArena Hard Prompts | 1414#141 | LMArena |
| GPQA Diamond (AA)not in index | 83.1%#158 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 24.5%#137 | Artificial Analysis |
Coding
| Benchmark | default | Source |
|---|---|---|
| LMArena Coding | 1451#126 | LMArena |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| Terminal-Bench Hard | 32.6%#95 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 94.4%#31 | Artificial Analysis |
Maths
| Benchmark | default | Source |
|---|---|---|
| AIME 2026 | 96.7%#6 | MathArena |
| HMMT February 2026 | 86.4%#20 | MathArena |
| MathArena Apex | 13.5%#18 | MathArena |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| AA-Omniscience | -41.8#263 | Artificial Analysis |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| IFBench | 66.5%#93 | Artificial Analysis |
Human preference
| Benchmark | default | Source |
|---|---|---|
| LMArena Text | 1394#138 | LMArena |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 63.0%#206 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| AA Intelligence Index | 17.0#203 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| SiliconFlow | 71 tok/s | 1.30 s | $0.100 | $0.300 | 262k | fp8 |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.020 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0001 | $0.0001 | 11.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0014 | $0.0007 | 13.1 s |
| Code edit | 6,000 / 1,500 | $0.0011 | $0.0007 | 17.2 s |
| Agentic coding session | 60,000 / 4,000 | $0.0072 | $0.0036 | 28.8 s |
| Structured extraction | 2,000 / 200 | $0.0003 | $0.0001 | 11.2 s |
See also
Data as of 9 Sept 2026. Compare with another model.