PeopleofSuperintelligence

Researcher

Yichao Zhang

Yichao Zhang is a DeepSeek-AI researcher whose work spans open language models, reinforcement learning and agent-training systems. He coauthored DeepSeek LLM and DeepSeek-V2 in 2024, DeepSeek-V3 in December 2024, and DeepSeek-R1 in 2025. DeepSeek-V3.2 introduced sparse attention in December 2025, and DeepSeek-V4 supports one-million-token contexts. In September 2026, he coauthored DeepSeek-V4.1-Flash on KV-cache compression and DeepSeek Elastic Compute (DSec) for agent training. As of October 2026, he contributes to DeepSeek research on language models and agent training.

Now at
DeepSeek

Timeline

  1. January 2024

    DeepSeek LLM

    In January 2024, he coauthors DeepSeek LLM, a project that uses scaling laws to advance open-source language models.

  2. May 2024

    DeepSeek-V2 efficiency

    In May 2024, he coauthors DeepSeek-V2, a mixture-of-experts model built for economical training and efficient inference. Compared with DeepSeek 67B, it saves 42.5% of training costs and cuts KV-cache use by 93.3%.

  3. December 2024

    DeepSeek-V3 report

    In December 2024, he coauthors the DeepSeek-V3 report, which describes a 671B-parameter mixture-of-experts model with 37B activated for each token.

  4. 2025

    DeepSeek-R1 reasoning

    In 2025, as part of the DeepSeek-AI Team, he coauthors DeepSeek-R1, showing that pure reinforcement learning can incentivize language-model reasoning without human-labelled reasoning trajectories.

  5. December 2025

    Sparse attention in V3.2

    In December 2025, he coauthors DeepSeek-V3.2, which introduces DeepSeek Sparse Attention to reduce computation while preserving performance on long-context tasks.

  6. April 2026

    DeepSeek-V4 context

    In April 2026, he coauthors the DeepSeek-V4 report on two mixture-of-experts models, both supporting contexts of one million tokens.

  7. September 2026

    KV-cache compression

    In September 2026, he coauthors DeepSeek-V4.1-Flash, a report on KV-cache compression that reduces the global cache footprint to 890 bytes per token.

  8. As of October 2026

    DeepSeek Elastic Compute

    In September 2026, he coauthors DeepSeek Elastic Compute (DSec), a production sandbox platform for agentic training at scale. As of October 2026, he contributes to DeepSeek research on language models and agent training.

  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 Yichao Zhang yet. Send us a free-licensed photo.

  1. DeepSeek LLM: Scaling Open-Source Language Models with Longtermism 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 | European Laboratory for Learning and Intelligent Systems ellis.eu. Checked 22 October 2026.
  5. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning | Nature nature.com. Checked 22 October 2026.
  6. DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 22 October 2026.
  7. DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 22 October 2026.
  8. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 22 October 2026.
  9. DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale arxiv.org. Checked 22 October 2026.