StepFun's Step 5 Preview: a large MoE model (reported ~600B total, 27B active, 1M context) for coding and reasoning.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Step 5 Preview is a preview release of StepFun's Step 5 language model, aimed at autonomous coding and reasoning tasks. StepFun (阶跃星辰) was founded in April 2023 and previously released the Step-1 language model and Step-1V multimodal model.
According to a YouTube creator review (AI Search), the model is a mixture-of-experts with roughly 600B total parameters, 27B active, and a 1M-token context window. It is described as one of the stronger open models, slightly behind GLM 5.3 and Mimo 2.6, costing about $1 per task via API, with weights around 650 GB. These details come from a secondary source and were not confirmed on the official page, which provided only a company description.
Access appears to be through StepFun's platform and API, with downloadable weights reported but not verified. The license, official documentation and exact release date were not found in the material provided.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
StepFun's flagship open-weights model for autonomous coding and reasoning, 600B MoE with 27B active and 1M context. Among the best open models, slightly behind GLM 5.3 and Mimo 2.6, at about $1 per task via API; 650 GB.
Multimodal agent MoE model with 600B total and 27B active parameters and 1M context. One point behind GLM 5.3 on Artificial Analysis but about three times cheaper; open release planned for October 15.
Closest neighbours by meaning, not just by category.
Model · StepFun
Open-weights multimodal flash model from StepFun for agentic tasks involving images, interfaces, web search and long runs.
Model · Alibaba
Alibaba's open-weights 2.4T-parameter MoE language model (95B active) from the Qwen3.8 family, with a hosted Max version.
Model · Z.ai
Z.ai's open-weights 753B MoE language model, a post-trained GLM-5.2 aimed at coding, agentic tasks and cybersecurity.
Model · Baidu
Baidu's open-weights 280B-parameter sparse MoE base model (14B active) with 1M-token context, released pre-post-training.
Model
Open model family (9B dense, 35B MoE, 397B MoE) trained with a self-improvement loop for reasoning, coding and agentic tasks.
Model · Meituan
Meituan's open-weights 1.6T-parameter multimodal MoE model (48B active) for agentic coding and visual tasks.