PeopleofSuperintelligence

Researcher

Xinyu Yang

Xinyu Yang is a DeepSeek researcher credited on its V2, V3, R1, V3.2, V4 and V4.1-Flash work. In 2024, he contributed to V2 and V3; V2-Lite is a 16-billion-parameter model with 27 layers and a 2,048-wide hidden dimension. He coauthored DeepSeek-R1 in 2025 and appeared in the V3.2 roster in December 2025. As of October 2026, he is listed in DeepSeek's Research & Engineering group for V4.1-Flash, a multimodal model with 552 billion backbone parameters.

Now at
DeepSeek
Works
8

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. May 2024

    DeepSeek-V2 and V2-Lite

    In May 2024, DeepSeek credits him in its Research & Engineering roster for V2, a 236-billion-parameter mixture-of-experts model with 21 billion active parameters per token, Multi-head Latent Attention, and a 128K-token context window. DeepSeek-V2-Lite has 16 billion parameters, 27 layers, and a 2,048-wide hidden dimension, and uses MLA and DeepSeekMoE.

  2. December 2024

    The DeepSeek-V3 report

    In December 2024, DeepSeek credits him in its Research & Engineering group for V3, a 671-billion-parameter mixture-of-experts model that activates 37 billion parameters per token. DeepSeek pre-trains V3 on 14.8 trillion tokens.

  3. January 2025–September 2025

    DeepSeek-R1 and reinforcement learning

    In January 2025, he coauthors DeepSeek-R1, which uses pure reinforcement learning to develop reasoning behaviors such as self-reflection, verification, and dynamic strategy adaptation. Nature publishes the study in September 2025 and lists him among the DeepSeek-AI authors.

  4. December 2025

    DeepSeek-V3.2 research

    In December 2025, DeepSeek lists him in the Research & Engineering roster for V3.2. The model introduces DeepSeek Sparse Attention and a scalable reinforcement-learning framework for reasoning and agent performance.

  5. April 2026

    DeepSeek-V4 context

    In April 2026, DeepSeek lists him in the Research & Engineering roster for V4. V4-Pro has 1.6 trillion parameters and V4-Flash has 284 billion parameters; both support a context length of one million tokens.

  6. As of October 2026

    DeepSeek-V4.1-Flash research

    As of October 2026, he is listed in DeepSeek's Research & Engineering roster for V4.1-Flash, a multimodal mixture-of-experts model with 552 billion backbone parameters and contexts up to one million tokens.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes DeepSeek. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes DeepSeek. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staffIncludes DeepSeek. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staff. We test the estimate against 3,913 named model contributors and dated all-staff records across Google DeepMind, OpenAI and Anthropic, plus six additional frontier-grade labs.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes DeepSeek. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Xinyu Yang yet. Send us a free-licensed photo.

  1. DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 22 October 2026.
  2. DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model arxiv.org. Checked 22 October 2026.
  3. DeepSeek-V3 Technical Report arxiv.org. Checked 22 October 2026.
  4. DeepSeek-V3 Technical Report arxiv.org. Checked 22 October 2026.
  5. DeepSeek-V3 Technical Report | European Laboratory for Learning and Intelligent Systems ellis.eu. Checked 22 October 2026.
  6. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 22 October 2026.
  7. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 22 October 2026.
  8. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning | Nature nature.com. Checked 22 October 2026.
  9. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 22 October 2026.
  10. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 22 October 2026.
  11. DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 22 October 2026.
  12. DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 22 October 2026.
  13. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 22 October 2026.
  14. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 22 October 2026.