Step 3.7 Flash

StepFun's multimodal Mixture-of-Experts model that processes text, images, and video with a 196B parameter backbone while activating approximately 11B parameters per inference.

Strengths

Efficiently handles multimodal inputs including video understanding while maintaining low computational overhead through sparse activation.

Best for

Applications requiring fast image and video analysis alongside text processing, particularly where latency and token efficiency matter.

Limitations

As a sparse model, it may not match dense models on complex reasoning tasks requiring sustained deep computation across very long contexts.