PeopleofSuperintelligence

Researcher

Xiaohan Wang

Xiaohan Wang is a DeepSeek-AI research and engineering contributor whose work spans language models, reinforcement-learning reasoning, and long-context efficiency. He coauthored DeepSeek-V2 and DeepSeek-V3, then DeepSeek-R1, which showed how pure reinforcement learning can develop language-model reasoning; Nature published the R1 article in 2025. His later papers introduced DeepSeek Sparse Attention, gave DeepSeek-V4 models million-token context lengths, and combined DeepSeek-V4.1-Flash with FP4 key-value caching. By September 2026, he is a Research & Engineering contributor at DeepSeek-AI.

Now at
DeepSeek

Timeline

  1. May 2024

    The DeepSeek-V2 model

    DeepSeek-V2's May 2024 report presents a mixture-of-experts language model with economical training and efficient inference, 236 billion total parameters and 21 billion activated per token; Wang is a coauthor.

  2. December 2024

    DeepSeek-V3's parameter scale

    DeepSeek-V3's December 2024 report describes a mixture-of-experts model with 671 billion total parameters and 37 billion activated per token; Wang is a coauthor.

  3. January 2025

    Reasoning with DeepSeek-R1

    DeepSeek-R1's January 2025 paper shows that pure reinforcement learning can incentivize reasoning abilities in language models, and Wang is a coauthor.

  4. September 2025

    DeepSeek-R1 in Nature

    In September 2025, Nature publishes the DeepSeek-R1 article with Wang among the DeepSeek-AI Team authors.

  5. December 2025

    DeepSeek Sparse Attention

    DeepSeek-V3.2 introduces DeepSeek Sparse Attention in December 2025, an attention mechanism that reduces computational complexity while preserving model performance in long-context scenarios; Wang coauthors the paper.

  6. April 2026

    Million-token DeepSeek-V4

    DeepSeek-V4's April 2026 report presents V4-Pro and V4-Flash with context lengths of one million tokens; Wang coauthors the report.

  7. September 2026

    DeepSeek-V4.1-Flash cache compression

    DeepSeek-V4.1-Flash combines cross-layer key-value cache reuse in Compressed Sparse Attention 2 with FP4 caching, giving the model an 890-byte-per-token global key-value cache footprint in September 2026; Wang coauthors the paper.

  8. September 2026

    Research at DeepSeek-AI

    Research & Engineering contributor at DeepSeek-AI is Wang's role as of September 2026.

  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.

Sources and credits

Photo: We have no published photo of Xiaohan Wang 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-V3 Technical Report arxiv.org. Checked 22 October 2026.
  3. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning arxiv.org. Checked 22 October 2026.
  4. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning nature.com. Checked 22 October 2026.
  5. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 22 October 2026.
  6. [2606.19348] DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 22 October 2026.
  7. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 22 October 2026.
  8. Xiaohan Wang | 500AI 500ai.org. Checked 22 October 2026.