PeopleofSuperintelligence

Researcher

Qiushi Du

Qiushi Du is a researcher whose work spans open language models, coding, theorem proving, and training infrastructure. He coauthored DeepSeek LLM in January 2024, DeepSeek-Coder-V2 in June 2024, and DeepSeek-Prover-V1.5 and Fire-Flyer in August 2024. His 2024 work also included DeepSeek-V3, and he coauthored DeepSeek-R1 and DeepSeek-V3.2 in 2025, the million-token DeepSeek-V4 preview in April 2026, and DSec in September 2026. He is a coauthor of the DeepSeek-AI group report DeepSeek-V4.1-Flash, a multimodal model with 552B backbone parameters published in September 2026.

Now at
DeepSeek
Works
28

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. January 2024

    DeepSeek LLM

    In January 2024, he coauthors DeepSeek LLM, a paper that studies scaling laws for open-source language models.

  2. June 2024

    DeepSeek-Coder-V2 code model

    In June 2024, he coauthors DeepSeek-Coder-V2, an open-source mixture-of-experts code model. It expands support from 86 to 338 programming languages and extends the context length from 16K to 128K.

  3. August 2024

    Lean theorem proving and AI-HPC

    In August 2024, he coauthors DeepSeek-Prover-V1.5, an open-source model for theorem proving in Lean 4 that builds on DeepSeek-Prover-V1. He also coauthors the Fire-Flyer AI-HPC paper, whose Fire-Flyer 2 system uses 10,000 PCIe A100 GPUs and approaches DGX-A100 performance while reducing costs by half.

  4. December 2024

    The DeepSeek-V3 report

    The December 2024 DeepSeek-V3 report names him among its authors and describes a mixture-of-experts model with 671B total parameters, 37B activated for each token.

  5. September 2025

    DeepSeek-R1 in Nature

    In September 2025, he coauthors DeepSeek-R1 as part of the DeepSeek-AI Team in Hangzhou, China. The paper shows that pure reinforcement learning can incentivize reasoning in large language models.

  6. December 2025

    DeepSeek-V3.2 and gold medals

    In December 2025, he coauthors DeepSeek-V3.2, which introduces DeepSeek Sparse Attention for long-context scenarios. The high-compute V3.2-Speciale achieves gold-medal performance at both the 2025 International Mathematical Olympiad and International Olympiad in Informatics.

  7. April 2026

    The DeepSeek-V4 preview

    The April 2026 DeepSeek-V4 preview includes him among its authors. It describes V4-Pro with 1.6T parameters (49B activated) and V4-Flash with 284B parameters (13B activated), both supporting a one-million-token context.

  8. September 2026

    DeepSeek-AI team coauthor

    In September 2026, he coauthors DeepSeek Elastic Compute (DSec), whose production-scale unit spans around 160 nodes and serves about 3 million sandboxes per day. He also coauthors DeepSeek-V4.1-Flash with the DeepSeek-AI author group; the multimodal mixture-of-experts model has 552B backbone parameters.

  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 Qiushi Du yet. Send us a free-licensed photo.

  1. [2401.02954] DeepSeek LLM: Scaling Open-Source Language Models with Longtermism arxiv.org. Checked 21 October 2026.
  2. [2406.11931] DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence arxiv.org. Checked 21 October 2026.
  3. [2408.08152] DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search arxiv.org. Checked 21 October 2026.
  4. DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search | Cool Papers - Immersive Paper Discovery papers.cool. Checked 21 October 2026.
  5. [2408.14158] Fire-Flyer AI-HPC: A Cost-Effective Software-Hardware Co-Design for Deep Learning arxiv.org. Checked 21 October 2026.
  6. [2412.19437] DeepSeek-V3 Technical Report arxiv.org. Checked 21 October 2026.
  7. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning | Nature nature.com. Checked 21 October 2026.
  8. [2512.02556] DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models arxiv.org. Checked 21 October 2026.
  9. [2606.19348] DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence arxiv.org. Checked 21 October 2026.
  10. [2609.22978] DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale arxiv.org. Checked 21 October 2026.
  11. [2609.19969] DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression arxiv.org. Checked 21 October 2026.
  12. DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression | ArXivSignals arxivsignals.io. Checked 21 October 2026.