StepFun
Chinese multimodal foundation model lab
AI nativeStepFun builds large multimodal models spanning text, voice, and video, and is one of China's prominent foundation model startups.
- Models
- 3
- Founded
- 2023
- AI native
- Yes
- Headquarters
- Shanghai, China
StepFun and AI
StepFun builds multimodal models for text, voice and video. Step 3.5 Flash came out in February 2026 under Apache 2.0, Step 3.7 Flash in May, and Step 5 Preview on September 18 with closed weights, a departure from the open releases before it.
The company has raised heavily and is preparing a Hong Kong IPO. It announced a Series B+ of over 5 billion yuan in January, and in May reportedly closed a $2.5 billion round while dismantling its offshore structure. One source says over 30 billion yuan, which conflicts. In June a report said it might file at a $12 billion valuation. We found no confirmation that the listing has happened.
// founder
- Jiang Daxin
AI models by StepFun 3
StepFun FAQ
- What is StepFun?
- StepFun builds large multimodal models spanning text, voice, and video, and is one of China's prominent foundation model startups.
- Who founded StepFun?
- StepFun was founded in 2023 by Jiang Daxin.
- Where is StepFun headquartered?
- StepFun is headquartered in Shanghai, China.
- Which AI models has StepFun released?
- StepFun has 3 models in our directory, including Step 5 Preview, Step-2 and Step-1.