StepFun, open weights
Step 3.5 Flash
Step 3.5 Flash by StepFun ranks 116th of 354 ranked models on the Noometry Index as of October 2026, with a score of 42.3. Its strongest category is math, where it ranks 84th. API pricing starts at $0.10 per million input tokens and $0.30 per million output tokens, with a 256K-token context window.
Last verified
Specifications
- Noometry rank
- #116 of 354
- Index score
- 42.3
- Evidence
- Confirmed 19 results
- Provider
StepFun- Released
- January 29, 2026
- Weights
- Open weights
- Reasoning
- Yes
- Context window
- 256K
- Max output
- 256K
- Input price
- $0.10 / M
- Output price
- $0.30 / M
- Blended price
- $0.15 / M
- Output speed
- Not measured
- Value
- #23 of 219
- Knowledge cutoff
- January 2025
- Input
- text
- Hugging Face
- stepfun-ai/Step-3.5-Flash
Category scores
Each category score combines every public result we have in that category.
- Coding 42.4
- Reasoning 22.2
- Math 42.6
- Knowledge 39.6
- Multilingual 50.5
- Instruction Following 73.1
- Long Context 42.8
- Writing & Preference 58.8
| Category | Score | Rank | Results |
|---|---|---|---|
| Coding | 42.4 | #105 | 1 |
| Reasoning | 22.2 | #202 | 2 |
| Math | 42.6 | #84 | 2 |
| Knowledge | 39.6 | #132 | 1 |
| Multilingual | 50.5 | #119 | 1 |
| Instruction Following | 73.1 | #124 | 1 |
| Long Context | 42.8 | #117 | 1 |
| Writing & Preference | 58.8 | #113 | 3 |
Strengths and weaknesses
Categories where Step 3.5 Flash places highest and lowest among the models ranked in each, with its score against that category's median.
Strongest categories
| Category | Score | vs median | Rank |
|---|---|---|---|
| Math | 42.6 | +6.1 | #84 of 327, top 26% |
| Coding | 42.4 | +3.6 | #105 of 340, top 31% |
| Writing & Preference | 58.8 | +5.0 | #113 of 312, top 37% |
Weakest categories
| Category | Score | vs median | Rank |
|---|---|---|---|
| Reasoning | 22.2 | −1.4 | #202 of 350, top 58% |
| Knowledge | 39.6 | +2.3 | #132 of 314, top 43% |
| Instruction Following | 73.1 | +1.8 | #124 of 305, top 41% |
Closest competitors
The models ranked just above and below Step 3.5 Flash. When scores are this close, price and speed are often the better way to choose.
| Model | Rank | Score | Blended $/M | Speed | |
|---|---|---|---|---|---|
| Qwen3.5-Flash | #112 | 42.5 | $0.18 | — | Compare |
| Nemotron 3 Ultra | #113 | 42.5 | $0.93 | — | Compare |
| Hunyuan T1 20250711 | #114 | 42.5 | — | — | Compare |
| DeepSeek-R1 | #115 | 42.3 | $0.91 | 10 | Compare |
| Qwen3.6 27B | #117 | 42.2 | $1.35 | — | Compare |
| Amazon Nova Experimental Chat 10 20 | #118 | 42.1 | — | — | Compare |
| Qwen3.5 122B-A10B | #119 | 42.1 | $1.10 | — | Compare |
| Longcat Flash Chat | #120 | 42.1 | — | 69 | Compare |
Sponsored placements are available on pages like this one. Advertise on Noometry
Benchmark results
Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.
Coding
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Coding | 1436 | #103 of 294, top 36% | LMArena | 2026-10-08 |
Reasoning
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| NYT Connections (extended) | 28.4% | #74 of 91, top 82% | Lech Mazur benchmarks | ||
| LMArena Hard Prompts | 1411 | #117 of 297, top 40% | LMArena | 2026-10-08 |
Math
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| MathArena Final-Answer Competitions | 66.8% | #19 of 29, top 66% | MathArena | ||
| LMArena Math | 1408 | #113 of 285, top 40% | LMArena | 2026-10-08 |
Knowledge
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Expert | 1421 | #105 of 273, top 39% | LMArena | 2026-10-08 |
Multilingual
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Non-English | 1385 | #119 of 297, top 41% | LMArena | 2026-10-08 | |
| LMArena Chinese | 1447 | #106 of 285, top 38% | LMArena | 2026-10-08 | |
| LMArena French | 1421 | #98 of 223, top 44% | LMArena | 2026-10-08 | |
| LMArena German | 1405 | #91 of 231, top 40% | LMArena | 2026-10-08 | |
| LMArena Japanese | 1354 | #94 of 211, top 45% | LMArena | 2026-10-08 | |
| LMArena Korean | 1352 | #104 of 213, top 49% | LMArena | 2026-10-08 | |
| LMArena Russian | 1385 | #124 of 283, top 44% | LMArena | 2026-10-08 | |
| LMArena Spanish | 1419 | #93 of 226, top 42% | LMArena | 2026-10-08 |
Instruction Following
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Instruction Following | 1385 | #119 of 298, top 40% | LMArena | 2026-10-08 |
Long Context
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Longer Query | 1402 | #115 of 291, top 40% | LMArena | 2026-10-08 |
Writing & Preference
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Text | 1403 | #117 of 297, top 40% | LMArena | 2026-10-08 | |
| LMArena Creative Writing | 1357 | #124 of 295, top 43% | LMArena | 2026-10-08 | |
| LMArena Multi-Turn | 1405 | #116 of 295, top 40% | LMArena | 2026-10-08 |
API pricing by provider
| Route | Input $/M | Output $/M | Cached input $/M | Checked |
|---|---|---|---|---|
| openrouter | $0.10 | $0.30 | — | 2026-10-10 |
| stepfun | $0.10 | $0.30 | $0.02 | 2026-10-10 |
Compare Step 3.5 Flash
- Step 3.5 Flash vs DeepSeek-R1
- Step 3.5 Flash vs Qwen3.6 27B
- Step 3.5 Flash vs Hunyuan T1 20250711
- Step 3.5 Flash vs Amazon Nova Experimental Chat 10 20
- Step 3.5 Flash vs Nemotron 3 Ultra
- Step 3.5 Flash vs Qwen3.5 122B-A10B
- Step 3.5 Flash vs GPT-6 Astra
- Step 3.5 Flash vs Claude Fable 5.1
- Step 3.5 Flash vs Gemini 3.8 Flash
- Step 3.5 Flash vs Kimi K3
- Step 3.5 Flash vs Grok 4.6
- Step 3.5 Flash vs Qwen3.8 Max
- Step 3.5 Flash vs GLM-5.3
- Step 3.5 Flash vs Muse Spark 1.3
Other StepFun models
Frequently asked questions
How good is Step 3.5 Flash?
Step 3.5 Flash by StepFun ranks 116th of 354 ranked models on the Noometry Index as of October 2026, with a score of 42.3. Its strongest category is math, where it ranks 84th. API pricing starts at $0.10 per million input tokens and $0.30 per million output tokens, with a 256K-token context window.
How much does Step 3.5 Flash cost?
Step 3.5 Flash costs $0.10 per million input tokens and $0.30 per million output tokens on StepFun's own API, with cached input at $0.02.
What is Step 3.5 Flash's context window?
Step 3.5 Flash accepts up to 256K tokens of input and can write up to 256K tokens in one response.
Is Step 3.5 Flash open source?
Yes. Step 3.5 Flash's weights are downloadable from Hugging Face (stepfun-ai/Step-3.5-Flash); check the license for commercial terms.
What are Step 3.5 Flash's strengths and weaknesses?
Relative to other ranked models, Step 3.5 Flash places best in math, coding, writing & preference and lowest in reasoning, knowledge, instruction following.
What is Step 3.5 Flash best at?
Its best category is math, where it ranks 84th on Noometry.