TLDRocket
Sign in

StepFun released an open-weight Step 5 Preview Mixture-of-Experts (MoE) model for coding and long-running agent workloads

Open source release Provisional 78% confidence first seen

StepFun launched Step 5 Preview, an open-weight Mixture-of-Experts model aimed at coding, financial research, and long-running agent tasks. The release includes checkpoints available for self-hosting and deployment support (e.g., vLLM, SGLang, Transformers), with reported specs of 600B total parameters, 27B active per token, and up to a 1M-token context. Checkpoints were published on platforms including Hugging Face and ModelScope, including FP8 and BF16 variants.

Step 5 Preview StepFun Hugging Face MarkTechPost ModelScope SGLang Transformers vLLM

Source coverage

Step 5 Preview Product Hunt social StepFun Releases Step 5 Preview Open Weights: 600B MoE, 27B Active, 1M Context MarkTechPost

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.