Step 3.5 Flash
StepFun · released Jan 29, 2026
Step 3.5 Flash is StepFun's most capable open-source foundation model. Built on a sparse Mixture of Experts (MoE) architecture, it selectively activates only 11B of its 196B parameters per token....
Specification
- Context window
- 262K
- Max output
- 66K
- Knowledge cutoff
- Not stated
- Parameters
- 199.4B
- Licence
- apache-2.0
- Serving providers
- 1
- Moderated
- No
- Uptime
- 100.0%
Intelligence
26.5
70th percentile
Coding
—
Coding Index
Agentic
—
Agentic Index
Output speed
56 t/s
Median across providers
Latency
1.39s
Time to first token
Cost per task
$0.03
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- τ²-bench (Telecom)87.4%
- GPQA Diamond82.6%
- IFBench66.5%
- AA-LCR (long context)60.3%
- SciCode38.5%
- Terminal-Bench Hard32.6%
- Humanity's Last Exam24.5%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| τ²-bench (Telecom) | 87.4% |
| GPQA Diamond | 82.6% |
| IFBench | 66.5% |
| AA-LCR (long context) | 60.3% |
| SciCode | 38.5% |
| Terminal-Bench Hard | 32.6% |
| Humanity's Last Exam | 24.5% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- Ling-2.6-1TInclusionAI26.6
- Step 3.5 FlashStepFun26.5
- o4 Mini HighOpenAI26.1
- o4 MiniOpenAI26.1
- Claude Opus 4Anthropic26.0
- Claude Sonnet 4Anthropic26.0
- Gemini 2.5 ProGoogle25.9
- GPT-5 MiniOpenAI25.8
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| Ling-2.6-1T | 26.6 |
| Step 3.5 Flash | 26.5 |
| o4 Mini High | 26.1 |
| o4 Mini | 26.1 |
| Claude Opus 4 | 26.0 |
| Claude Sonnet 4 | 26.0 |
| Gemini 2.5 Pro | 25.9 |
| GPT-5 Mini | 25.8 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.1
- Output / 1M tokens
- $0.3
- Cached input / 1M
- Not offered
- Blended 3:1
- $0.15
One task, estimated
$0.03
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.