Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model. It pairs a 196B-parameter language backbone with a vision encoder for native image and video understanding, activating roughly 11B parameters...
Details
| Field | Value |
|---|---|
| id | stepfun/step-3.7-flash |
| modality | text+image+video->text |
| context_length | 256000 |
| max_input_tokens | 256000 |
| max_output_tokens | 230400 |
| knowledge_cutoff | 2026-03-01 |
| release_date | 2026-05-29 |
| is_open_weights | yes |
| supports_tool_call | yes |
| supports_reasoning | yes |
| supports_structured_output | yes |
| supports_attachment | yes |
| input_modalities | text, image, video |
| output_modalities | text |