Step 3.7 Flash
Step 3.7 Flash adds native visual input to StepFun's fast reasoning line. Its 198B total parameters include a 196B language backbone and a 1.8B vision encoder. About 11B parameters activate per token.
Keep visual evidence in the task
The API can process images and video without a separate visual model. The provider positions this release for repeated tool use, coding and analysis of visual information.StepFun step-3.7-flash guide
Check the smaller visual model too
Practical guidance. Use this release when a coding or analysis task also needs screenshots or video. Its active parameter count describes computation per token; it does not mean that the full weights occupy only 11B parameters of memory.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable |
| Model identifiers | step-3.7-flashStep-3.7-Flash-FP8Step-3.7-Flash-NVFP4Step-3.7-Flash-GGUF |
| Context window | 256,000 tokens Provider-listed 256K context window. StepFun step-3.7-flash guide |
| License | Apache 2.0 for the published release.StepFun Step-3.7-Flash model card |
| Reasoning controls | Low, medium and high reasoning levels. The API accepts image and video input.StepFun step-3.7-flash guide |
| API formats | Chat Completions and Messages endpoints are documented.StepFun step-3.7-flash guide |
| Weight formats | FP8, NVFP4 and GGUF repositories are packaging variants of this release.StepFun official published model repositories |
Related models
- Step 5 Preview
StepFun · Step 5
A multimodal mixture-of-experts model with 600B total and 27B active parameters for text, image and video tasks.
Launch date not confirmedPreviewAPI · Downloadable · Provider product - Step 3.5 Flash
StepFun · Step 3
Step 3.5 Flash is a text reasoning model for code and tool use.
Launch date not confirmedAPI · Downloadable - Step3-VL-10B
StepFun · Step 3
Step3-VL-10B combines a 1.8B PE-lang visual encoder with a Qwen3-8B decoder.
Launch date not confirmedDownloadable