←── back to feed
/topics/step-3-7-flash-vision-language-model-release

Step 3.7 Flash vision-language model release

3 items2 sourcesupdated 16d agotrend 0

StepFun released Step 3.7 Flash, a 198B-parameter mixture-of-experts vision-language model with native vision capabilities, 256k context window, and Advisor Mode designed for coding agents and search workflows.

  • 198B-parameter MoE architecture with A11B active parameters
  • Native vision capabilities integrated into the model
  • 256k token context window
  • Advisor Mode feature for enhanced functionality
  • Optimized for coding agents and search workflows