Citation and evidence

Jiang Daxin

9 min full readUpdated 21 references

This article's verification

Report a problem with this article

More

Use this article

Raw MarkdownExplore connections

Improve this page

Suggest editRevision historyDiscussion

Browse categories

Chinese AIPeople

Cite this article

Jiang Daxin (Chinese: 姜大昕) is a Chinese computer scientist and technology entrepreneur who is the founder and chief executive officer of StepFun (Chinese: 阶跃星辰), a Shanghai-based artificial intelligence company that develops the Step family of language and multimodal foundation models. Before starting the company in April 2023, he spent about 16 years at Microsoft, where he became a Microsoft global partner and the deputy head and chief scientist of the Software Technology Center Asia (STCA), and led natural language processing and search work for products including the Bing search engine and the Cortana voice assistant. [4][5][14]

StepFun is counted among China's so-called "AI tigers," a cohort of well-funded large-model startups. [1] Under Jiang the company has bet on scaling laws and on models that handle text, images, audio, and video, and it has released a growing number of them as open weights. In January 2026 it announced a Series B+ round of more than 5 billion yuan and named Megvii co-founder Yin Qi as chairman, and on September 20, 2026 it introduced Step 5 Preview, a 600-billion-parameter mixture-of-experts model. [6][9][17]

Education

Jiang earned a PhD in computer science from the University at Buffalo, part of the State University of New York. [4][5] His research interests spanned machine learning, data mining, natural language processing, and bioinformatics. [4][5] After completing his doctorate he began his academic career in Singapore as an assistant professor in the School of Computer Science and Engineering at Nanyang Technological University. [4][5]

Microsoft career

In 2007 Jiang joined Microsoft Research Asia (MSRA) in Beijing as a principal researcher, beginning a Microsoft career that lasted about 16 years. [4][5] His work there spanned machine learning, data mining, and language technologies, and in 2008 he received the ACM SIGKDD Best Application Paper Award. [4][5]

In 2011 Jiang moved from MSRA to the Software Technology Center Asia, Microsoft's engineering organization in China. [5][14] In 2017 he became a Microsoft global partner, serving as deputy head and chief scientist of STCA, and in March 2023 he was promoted to vice president, according to the Chinese technology outlet Leiphone. [5] Before he left, his team had grown to more than 400 applied scientists and engineers working on search and AI algorithms, applications, and platforms. It was responsible for Bing's global search experience as well as natural language processing research and development. [5] Leiphone credits the team with extending Bing's intelligent question-answering feature in 2019 from about a dozen languages to more than 100 languages and 200 regions. [5] Baidu Baike describes his responsibilities as natural language understanding for Bing, Cortana, and Microsoft Azure. [4]

Over his career Jiang has published close to 200 papers in international conferences and journals. [4] He was named to the 2025 class of IEEE Fellows, announced in December 2024. [4]

StepFun

Jiang was one of several senior researchers who left large technology companies in 2023 to start large-model companies. [5] StepFun was registered on April 6, 2023, with Jiang as executive director and manager. [5] Its legal name is Shanghai Jieyue Xingchen Intelligent Technology Co., Ltd., and it uses StepFun in English. [3] Its core founding team included Zhu Yibo, who led systems, and Jiao Binxing, who led data; both had worked at Microsoft. [18][19] Zhu describes himself as a StepFun co-founder, and a February 2025 profile carried by Sina Finance names Jiao as a co-founder. [20][19] Zhu, described by the South China Morning Post in 2024 as the company's head of systems, later became chief technology officer. When Yin Qi was named chairman in January 2026, KrASIA and Yicai Global listed the management team as CEO Jiang, chief scientist Zhang Xiangyu and CTO Zhu. [2][9][21]

StepFun is grouped with China's "AI tigers," alongside rivals such as Zhipu AI, MiniMax, and Moonshot AI. [1][3][6] The company has publicly tied its strategy to the scaling-law hypothesis, the idea that model performance improves predictably as model size, data, and compute grow. [2] At the July 2024 World Artificial Intelligence Conference, Jiang said: "Climbing the peak of AGI, 'trillion parameters' and 'multi-modal fusion' are indispensable. The scale of trillion parameters is the basic threshold for achieving AGI; multi-modal large models are the only way to AGI." [10] In a May 2025 interview summarized by Baidu Baike, he described two trends in the field: a shift from imitation learning to reinforcement learning, and a move from combining modalities toward models that integrate multimodal understanding and generation. [4]

Models

By June 2024 StepFun had launched Step-1V, a multimodal model with more than 100 billion parameters, and was testing a model with more than one trillion parameters. [2] At the July 2024 World Artificial Intelligence Conference it released the official version of Step-2, a trillion-parameter language model built on a mixture-of-experts architecture, together with the Step-1.5V multimodal model and the Step-1X image generator. [10] A November 2024 MarkTechPost report said Step-2 ranked fifth on the LiveBench benchmark. [8]

The company then expanded across modalities. In February 2025 StepFun and Geely Auto announced that they would open-source the Step-Video-T2V text-to-video model, which generates 204-frame videos at 540p resolution, and the Step-Audio speech model. [14] Its flagship Step-3, released in July 2025, is a multimodal reasoning model built on a mixture-of-experts architecture with 321 billion total parameters and 38 billion active parameters. StepFun says it was designed through the co-design of two techniques it calls Multi-Matrix Factorization Attention (MFA) and Attention-FFN Disaggregation (AFD), and it published the weights under the Apache 2.0 license. [11] The company has since released smaller open-weight models under Apache 2.0, including Step 3.5 Flash in February 2026 and Step 3.7 Flash in May 2026. [12][13] On September 20, 2026, StepFun introduced Step 5 Preview as its new flagship for agentic work, with API access at launch and open weights promised for October 15. [17]

ModelReleasedNotes
Step-1VBy June 2024Multimodal model with more than 100 billion parameters [2]
Step-2July 2024Trillion-parameter MoE language model (official version) [10]
Step-1.5V / Step-1XJuly 2024Multimodal model and image generator [10]
Step-Video-T2V / Step-AudioFebruary 2025Open-sourced with Geely; text-to-video and speech models [14]
Step-3July 2025321B-parameter MoE (38B active); MFA and AFD; Apache 2.0 [11]
Step 3.5 FlashFebruary 2026196B parameters, 11B active per token; 256K context; Apache 2.0 [12]
Step 3.7 FlashMay 2026196B language backbone plus 1.8B vision encoder, about 11B active; Apache 2.0 [13][16]
Step 5 PreviewSeptember 2026600B total, 27B active MoE; 1M context and vision input; weights promised for October 15, 2026 [17]

Expanded article table

Funding and business

StepFun has raised capital from Tencent, venture investors, and Shanghai state-backed funds. A Series B round disclosed in December 2024 was led by Fortera Capital, the private equity arm of Shanghai State-owned Capital Investment, with Tencent and Qiming Venture Partners also participating; the amount was not disclosed but was reported as several hundred million dollars. [7] On January 26, 2026, the company announced a Series B+ round of more than 5 billion yuan (about 717 million dollars), which Caixin reported was larger than the proceeds of the Hong Kong listings that Zhipu AI and MiniMax had completed earlier that month. [6] New investors in that round included the Shanghai State-owned Capital Investment Leading Fund, China Life Private Equity Investment, Pudong Venture Capital, Xuhui Capital, Xiamen ITG Group, and Huaqin Technology, while Tencent, Qiming Venture Partners, and 5Y Capital returned. [9] At the same time, the company named Yin Qi, co-founder of Megvii, as chairman. [6][9]

Rather than rely on a single consumer chatbot, StepFun has pushed its models into devices and vehicles. Caixin reported partnerships with the smartphone makers Honor, Oppo, and ZTE. [6] In the automotive sector, Geely's Agent OS system, installed in models such as the Galaxy M9, is powered by StepFun's end-to-end voice model, and StepFun said it expected its models to be deployed in more than one million vehicles in 2026. [9] In July 2025, StepFun launched an alliance with Chinese AI chipmakers including Huawei Ascend, Cambricon, Biren, and Moore Threads, aimed at optimizing Step-3 for domestic chips. [15]

In June 2026, Chinese financial media reported, citing people familiar with the matter, that StepFun had confidentially filed for a Hong Kong initial public offering and that its main investors had proposed a valuation of up to 12 billion dollars. The same report said the company had completed a shareholding reform in April 2026 and raised nearly 2.5 billion dollars in a pre-IPO round in May from investors including Huaqin Technology, Longcheer, OmniVision, ZTE, and the Hong Kong Investment Corporation. [16] StepFun had not publicly confirmed a filing in that report, and the figures are unnamed-source estimates.

Recognition

  • Fellow of the IEEE, 2025 class (announced December 2024). [4]
  • Named one of the 10 "AI Figures of the Year 2025" in the 2025 China AI Awards, announced in July 2025. [4]
  • ACM SIGKDD Best Application Paper Award, 2008. [4][5]

References

  1. ^1 ^2The Wire China, "What is StepFun?" June 8, 2025. thewirechina.com/...what-is-stepfun
  2. ^1 ^2 ^3 ^4South China Morning Post, "Shanghai AI start-up founded by ex-Microsoft engineers bets on 'scaling law' to boost AI capabilities," June 2024. scmp.com/...bets-scaling-law-boost-ai-capabilities
  3. ^1 ^2Wikipedia, "StepFun" (accessed September 2026). en.wikipedia.org/...StepFun
  4. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13Baidu Baike (English edition), "Jiang Daxin (CEO of Shanghai Step Star Technology Co., Ltd.)." baike.baidu.com/...45936
  5. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13Leiphone (雷峰网), "前微软 NLP 大牛姜大昕创立新公司「阶跃星辰」," December 2023. leiphone.com/...mJcRRqbgIh6UkVuk
  6. ^1 ^2 ^3 ^4 ^5Caixin Global, "StepFun Raises $717 Million, Outpacing Newly Listed AI Rivals," January 26, 2026. caixinglobal.com/...wly-listed-ai-rivals-102408206
  7. ^SiliconANGLE, "Chinese AI model maker Stepfun raises hundreds of millions in Series B funding," December 26, 2024. siliconangle.com/...reds-millions-series-b-funding
  8. ^MarkTechPost, "Chinese AGI Startup 'StepFun' Developed 'Step-2': A New Trillion-Parameter MoE Architecture Model Ranking 5th on Livebench," November 20, 2024. marktechpost.com/...model-ranking-5th-on-livebench
  9. ^1 ^2 ^3 ^4 ^5KrASIA, "China's investors double down on AI frontrunners as StepFun raises RMB 5 billion," January 2026. kr-asia.com/...ers-as-stepfun-raises-rmb-5-billion
  10. ^1 ^2 ^3 ^4Pandaily, "StepFun Releases Three Large Models of the Step Series," July 2024. pandaily.com/...ee-large-models-of-the-step-series
  11. ^1 ^2StepFun, "Step3" model card, Hugging Face, July 2025. huggingface.co/...step3
  12. ^1 ^2StepFun, "Step-3.5-Flash" model card, Hugging Face, February 2026. huggingface.co/...Step-3.5-Flash
  13. ^1 ^2StepFun, "Step-3.7-Flash" model card, Hugging Face, May 2026. huggingface.co/...Step-3.7-Flash
  14. ^1 ^2 ^3 ^4Yicai Global, "China's Geely and Stepfun Join Open-Source AI Trend With Two Models," February 18, 2025. yicaiglobal.com/...source-ai-trend-with-two-models
  15. ^TechNode, "Chinese AI chipmakers join forces with StepFun to counter Nvidia's return to China," July 30, 2025. technode.com/...to-counter-nvidias-return-to-china
  16. ^1 ^2Sina Finance, "AI'六小虎'之一阶跃星辰,据传已秘密递表港交所,估值达120亿美元," June 10, 2026. finance.sina.com.cn/...doc-iniaxkrt1220200.shtml
  17. ^1 ^2 ^3StepFun (@StepFun_ai), "Introducing Step 5 Preview: Advancing the Pareto Frontier," X, September 20, 2026. x.com/...2101510462685003786
  18. ^Sina Finance (新浪财经), "最神秘的大模型独角兽,刚刚融了几亿美元" (The most mysterious large-model unicorn just raised several hundred million dollars), December 24, 2024. finance.sina.com.cn/...doc-ineapwes8461488.shtml
  19. ^1 ^2Sina Finance (新浪财经), "和DeepSeek一样强的公司,上海也潜伏一家" (Shanghai also has a company as strong as DeepSeek), February 16, 2025. finance.sina.com.cn/...doc-ineksafe5136606.shtml
  20. ^Yibo Zhu, personal homepage (accessed September 2026). yibozhu.com
  21. ^Yicai Global, "Chinese AI Firm Stepfun Raises USD719 Mln for Model Development, AI Agent Rollout," January 26, 2026. yicaiglobal.com/...development-terminal-agents-use

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

5 revisions · v6 · 1,843 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent verification 2026-09-23 (xg05 V14): IEEE class, chip alliance, Step 5 Preview checked

Cite this page: AI Wiki. "Jiang Daxin." aiwiki.ai, updated 23 Sept 2026, fact-checked 23 Sept 2026. CC BY 4.0. https://aiwiki.ai/wiki/jiang_daxin

Suggest edit