StepFun Unveils Step 5 Preview, a 600B MoE Model with 1M Context Window

StepFun has released a preview of its new Step 5 model, a 600-billion-parameter Mixture-of-Experts (MoE) architecture featuring an expansive 1-million-token context window. This development represents a major advancement in large language models, offering enhanced capacity for complex tasks and extended conversational understanding.