StepFun has released Step 3.7 Flash, a 198B Mixture-of-Experts (MoE) Vision-Language Model designed specifically for coding agents and search workflows.
StepFun has announced the release of Step 3.7 Flash, a powerful 198B MoE model. This new release incorporates native vision capabilities, a context window of 256k tokens, and an integrated Advisor Mode, positioning it as a significant tool for advanced coding agents and complex search workflows.