Step 3.7 Flash
Unavailable in this round
Language modelOpen source
Step 3.7 Flash
StepFun
- Released
- May 2026
- Data date
- September 8, 2026
- Predecessor
- Step 3.5 Flash
- Speed
- 81.1 tokens/sOutput throughput
- Price
- Input · Output / 1M tokens
- Input
- Text, Image
- Output
- Text
New in Step 3.7 Flash is image understanding in StepFun's fast Step line. The sparse MoE accepts text and images, returns text, activates about 11 billion parameters, and combines a 256K context with selectable low, medium, or high reasoning effort.
StepFun compares the release directly with Step 3.5 Flash and reports +5% on SWE-Bench Pro plus 6.1% on Terminal-Bench 2.1 in the launch text.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 198BSource |
| Active parameters | 11BSource |
| Context window | 256,000 tokensSource |
| Notes | Apache-2.0 multimodal MoE with 198B total parameters, about 11B active parameters, and a 256K context window.Source |
| License | Apache-2.0Source |
| Parameters | 198 BSource |
| Active parameters | 11 BSource |
| Architecture | sparse MoE vision-language model, 196B language backbone plus 1.8B vision encoderSource |
| Input | Text, Image (The checked official API/model-page material documents text and image inputs; video is not documented for this checkpoint.)Source |
| Output | TextSource |
| Reasoning mode | selectable low, medium and high reasoning levelsSource |
| Tool calling | YesSource |
| API model ID | step-3.7-flashSource |
| Throughput in tokens per second | 400 (Throughput is a usage claim, not a benchmark score.)Source |
| Release | 2026-05-29Source |
Page 1 of 2
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
SimpleVQA (Search) (Show measurement, test conditions, and source)
- Source value
- 79.2
- Score
- 79.2
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
V* (Python) (Show measurement, test conditions, and source)
- Source value
- 95.3
- Score
- 95.3
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
ClawEval-1.1 (Show measurement, test conditions, and source)
- Source value
- 67.1
- Score
- 67.1
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Toolathlon (Show measurement, test conditions, and source)
- Source value
- 49.5
- Score
- 49.5
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
HLE (with tools) (Show measurement, test conditions, and source)
- Source value
- 48.1
- Score
- 48.1
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
SWE-Bench Pro (Show measurement, test conditions, and source)
- Source value
- 56.3
- Score
- 56.3
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Terminal-Bench 2.1 (Show measurement, test conditions, and source)
- Source value
- 59.5
- Score
- 59.5
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
GDPVal-AA (Show measurement, test conditions, and source)
- Source value
- 45.8
- Score
- 45.8
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Step-SWE-Bench (Hermes) (Show measurement, test conditions, and source)
- Source value
- 67.5
- Score
- 67.5
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Step-SWE-Bench (OpenClaw) (Show measurement, test conditions, and source)
- Source value
- 67
- Score
- 67
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Step-SWE-Bench (Claude Code) (Show measurement, test conditions, and source)
- Source value
- 71.5
- Score
- 71.5
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Step-SWE-Bench (Kilo) (Show measurement, test conditions, and source)
- Source value
- 67.5
- Score
- 67.5
- Metric
- percent
- Unit
- %
- Category
- mixed
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- StepFun Step 3.7 Flash official model page
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official page/benchmark claims; preserve harness, precision and MTP configuration in row names.
Page 1 of 2
Our tests
SVG short test
One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.
Earlier tests
Pelican on a bicycle
A community classic for free-form SVG drawing.
Step 3.7 Flash
Reasoning (requested): high
Current attempt
- Run date
- Oct 1, 2026
- API provider
- openrouter
- Requested API model
- stepfun/step-3.7-flash
- API model reported in response
- stepfun/step-3.7-flash
- Reasoning setting
- high
- Pinned endpoint
- Novita | stepfun/step-3.7-flash-20260528
- Temperature
- Provider default
- Token limit including reasoning
- 8,192 tokens
- Test protocol
- community-visual-tests-v1
- Technical checks
- Passed, not a quality rating
- Response time
- 46.6 s
Visually reviewed. This individual test does not establish overall model quality.
Prompt and test conditions
Generate an SVG of a pelican riding a bicycle
Token limit including reasoning: 8,192 tokens
Task origin (Simon Willison)Popularity
41.6 / 100
Predecessor and successor
Sources and data date
| Type | Evidence and data date |
|---|---|
| Speed | StepFun. Output speed after the first streaming chunk, not total response time. Retrieved September 8, 2026. |
| Price | Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. |
| Input | Documented modalities. Retrieved September 8, 2026. |
| Output | Documented modalities. Retrieved September 8, 2026. |
| Research status | Research date September 8, 2026. 3 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | knowledge-cutoff Original wording The exact Step 3.7 Flash page does not publish a knowledge cutoff. |
| Additional source | Model metadata (retrieved September 8, 2026) · StepFun Step 3.7 Flash repository · Editorial description reviewed October 2, 2026 |
| Additional source | StepFun Step 3.5 Flash repository · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |