- 4Releases tracked
- Sep 20Latest · Step 5 Preview
- 75%Open-weight
- CNShanghai, China
StepFun is a Shanghai, China-based AI company that builds the Step family of foundation models for language, vision, audio and agentic work. Most of its tracked releases are open-weight on Hugging Face and GitHub: the multimodal Step 3 (July 2025), Step 3.5 Flash (February 2026), a 196B-total, 11B-active mixture-of-experts reasoning model with a 256K context window, and the multimodal Step 3.7 Flash (May 2026). Its flagship Step 5 Preview, announced September 20, 2026, is a 600B mixture-of-experts model with 27B active parameters, vision and a 1M-token context, aimed at agentic coding and knowledge work. StepFun also serves its models through an API platform.
StepFun
4 releases latest Sep 20, 2026
-
Step 5 Preview
600B MoE (27B active) flagship for agentic coding and knowledge work with vision and 1M context; open weights planned Oct 15.
-
Step 3.7 Flash
-
Step 3.5 Flash
-
Step 3
StepFun FAQ
- What is StepFun's latest AI model?
- StepFun's most recent release in the Startrise tracker is Step 5 Preview, released Sep 20, 2026.
- How many AI models has StepFun released?
- Startrise tracks 4 notable StepFun releases since January 2025, from Step 3 (Jul 31, 2025) to Step 5 Preview.
- Does StepFun release open-weight models?
- Yes. 3 of 4 tracked StepFun releases (75%) ship with downloadable weights; the most recent is Step 3.7 Flash.
- Where is StepFun based?
- StepFun is based in Shanghai, China.