Renaissance Labs

Atlas / Asia

Independent company · Shanghai, China

StepFun

StepFun is a Shanghai foundation-model company that has released open-weight language, speech and video models under permissive licenses.

language modelsmultimodalspeech

Shanghai, China

Founded
2023
Type
Independent company
Based in
Shanghai, China
Key papers
3
Featured projects
8

About

StepFun, formally Shanghai Jieyue Xingchen Intelligent Technology, was founded in April 2023 by Jiang Daxin, Zhu Yibo and Jiao Binxing, all formerly of Microsoft. It is often counted among China's six leading foundation-model startups, and its investors include Tencent and Qiming Venture Partners.

The company builds large language, multimodal, image, video and audio models. Step-2, a trillion-parameter model, appeared in 2024, and it open-sourced Step-Video-T2V and Step-Audio in February 2025 together with Geely.

Its recent open releases, Step 3.5 Flash in February 2026 and Step 3.7 Flash in May 2026, use the Apache 2.0 license. Its latest preview model, Step 5, is closed-weight, so openness varies by release.

Mission and ambitions

StepFun states an aim of building frontier multimodal models and making many of them available to developers.

Known for

  1. 01

    Open Step-Video and Step-Audio models

  2. 02

    Step 3.5 Flash mixture-of-experts model

  3. 03

    Trillion-parameter Step-2

Milestones

  1. 2024

    Step-2 and multimodal models

    StepFun announces the trillion-parameter Step-2, Step-1.5V and Step-1X.

  2. 2025

    Step-Video-T2V and Step-Audio

    Both are open-sourced in February 2025.

  3. 2025

    Step 3

    The company releases Step 3 in July 2025.

  4. 2026

    Step 3.5 Flash

    A 196B-parameter mixture-of-experts model with 11B active parameters is released under Apache 2.0.

  5. 2026

    Step 3.7 Flash

    A further Apache 2.0 release follows in May 2026.

Open source

GitHubgithub.com/stepfun-ai
Hugging Facehuggingface.co/stepfun-ai

Research papers

Notable releases

Openness

Weights released on Hugging Face and code on GitHub; licence per model.

People

Listed from the lab's own pages and public profiles.

Contact and community

Contact via the official site, GitHub issues or Hugging Face discussions.

Sources

Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.

Labs nearby