PJLab
academicThe Shanghai Artificial Intelligence Laboratory, officially known as the Pujiang Laboratory (浦江实验室), is a state-backed new-type research institution (新型研发机构) established in July 2020 by the Shanghai Municipal Government. The institution is most commonly referred to as PJLab in technical contexts (reflected in its domain, pjlab.org.cn) and SAIL in international media reports. Established in partnership with SJTU, Fudan University, and SenseTime, it serves as a core National Laboratory anchoring China's strategic "National Team" in AI.
With an estimated annual budget of 5–8 billion yuan ($700 million–$1.1 billion), it is heavily subsidized by the Shanghai AI Pilot Fund. The lab operates a dedicated compute cluster in the Lingang New Area housing over 30,000 accelerators. The lab maintains a symbiotic relationship with SenseTime, leveraging its SenseCore AI infrastructure for large-scale training. This collaboration produced the InternLM series and InternVL, which consistently lead open-source benchmarks.
Beyond traditional LLMs, the lab is a leader in "AI for Science" through the Shanghai Academy of AI for Science (SAIS). PJLab sits alongside BAAI (Beijing) and Pengcheng Lab (Shenzhen) as the three regional pillars of China's national AI infrastructure under the State Council's New Generation AI strategy.
Your notes
People
- Zhou Bowen (周伯文) Tsinghua FacultyGoogle ScholarLinkedIn — Director & Chief Scientist, Shanghai AI Laboratory (formerly SVP & President of JD Cloud & AI, JD.com; Chief Scientist, IBM Watson Group; Director, IBM AI Foundations Labs; IBM Distinguished Engineer; PhD University of Colorado Boulder; MS Chinese Academy of Sciences; BS USTC)
- Yu Qiao (乔宇) Google ScholarOpenReview — Lead Scientist, Shanghai AI Laboratory (formerly Professor & Director, SIAT/CAS; Postdoc, CUHK MMLab (2006-2007); PhD University of Electro-Communications Tokyo; BE/ME Chongqing University)
- Dahua Lin (林达华) Google ScholarOpenReviewCUHK Faculty — Senior Researcher & OpenMMLab Director (formerly Co-founder, SenseTime (2014); Associate Professor, CUHK; PhD MIT (2012); MPhil CUHK (2006); BEng USTC (2004))
- Jifeng Dai OpenReview — Senior Researcher & InternVL Lead (formerly Known for Deformable ConvNets)
- Wenhai Wang OpenReview — Core Researcher (InternImage, InternVL)
- Xizhou Zhu OpenReview — Core Researcher (formerly Known for deformable attention mechanisms)
News
- 2026-09-13 Intern-S2-397B Release Weights Published Under Apache 2.0, Succeeding the July Preview — Hugging Face
- 2026-09-11 Atria Dawn Preview Released: Agentic Model Post-Trained From GLM-5.2 (744B MoE, MIT, 256K) With a 143-Author Paper on a Verifiable Experience Pipeline — Hugging Face
- 2026-09-09 NCP-ArchPreview: 8.9B Latent-Space Language Model Trained on 5.73T Tokens With Next Concept Prediction; Reaches OLMo-3-7B's Loss at 51% of the Tokens — arXiv
- 2026-09-08 SWE-Bench Pro Verified: OpenCompass Ships an Anti-Hacking, Task-Corrected Version of the Repository-Level Coding Benchmark — arXiv
- 2026-09-04 SciDocBench: Workflow-Centered Scientific-Document Benchmark (124 Expert Questions, 496 Instances) With SFT and RL Data — arXiv
- 2024-12-13 Profile: Shanghai AI Lab: Driving both AI safety and development — MERICS