StepFun Releases Step 5 Preview Open Weights: 600B MoE, 27B Active, 1M Context
StepFun has released the Step 5 Preview model, an open-weights sparse Mixture-of-Experts architecture featuring 600 billion total parameters and 27 billion active parameters per token. Designed for long-horizon agentic tasks, the model supports a 1 million token context window. Checkpoints in FP8 and BF16 formats are now available for public download on Hugging Face and ModelScope.
ModelsStep 5 Preview
Covered by 1 source
- MMarkTechPost↗Michal Sutter2d ago