Shanghai Artificial Intelligence Laboratory is a new type of research institution in China's artificial intelligence field. Its official website takes "open, controllable, and trustworthy AGI" as its vision, proposing to build general artificial intelligence that integrates general and specialized capabilities, is open and shared, and is safe and trustworthy.
【Main Research Directions and Platforms】 The scientific research achievements and platforms listed on the official website include: Intern·Duanyan, Intern Science Discovery Platform, Intern S2, Intern Science Multimodal Large Model, Intern W0, Intern Physical World Model, Intern Safety, Safety Foundation Platform, SiNan, Large Model Open-Source Open Evaluation System, OpenDataLab, and Artificial Intelligence Open Data Platform. Related release information shows that the Intern·Duanyan Science Discovery Platform has fully opened its full-stack services, emphasizing the integration of "computing, proving, and verifying"; Intern S2 is described as the open-source foundation large model that best understands science, with scientific capabilities comparable to top closed-source models and general performance in the first tier of open source; Intern W0 is described as a physical world model, providing key support for AI to enter the physical world. The official website also mentions release information such as the Intern·Mingjue decision model and Intern SCP-Lab.
【Open Source and Openness】 The official website has an open source and openness section. The types of open-source projects listed include models, tools, algorithms, datasets, and evaluation sets. The projects displayed are: InternVL (2024.05.09, 9.4k), MinerU (2025.11.06, 48.7k), LMDeploy (2025.10.30, 7.2k), InternLM (2024.01.02, 7.1k), OpenCompass (2025.10.20, 13k), XTuner (2025.07.11, 5k).
【Technical Philosophy】 The official website explains that the integration of general and specialized capabilities is the strategic path to AGI, and describes that the integrated general-specialized technology system includes three mutually dependent levels: the foundation model layer, the integration and collaboration layer, and the exploration and evolution layer. In terms of embodied intelligence and computing power, it mentions the "Intern" embodied full-stack engine Intern-Robotics, as well as the DeepLink ultra-large-scale cross-domain mixed training technical solution, which achieves cross-domain long-term stable mixed training of hundred-billion-parameter large models across multiple intelligent computing centers over a thousand kilometers, and provides a software-hardware collaborative solution through "cloud-edge-device integrated training and inference." In terms of safety and trustworthiness, it proposes exploring the AI-45° Law technical system, taking causality as the core path, and exploring trustworthy AGI through three progressive stages: generalized alignment, intervenability, and reflectivity.
【Updates, Activities, and Talent】 The official website provides scientific research updates, news updates, and scientific research activity information, and mentions the 2026 Pujiang Innovation Forum, WAIC 2026Science Frontier Forum and other activities. In terms of talent, the official website carries out both social recruitment and campus recruitment, with directions including artificial intelligence, large models, embodied intelligence, safe and trustworthy AI, AI for Science, foundational platforms, innovation chains, management support, and more; since 2022, the laboratory has jointly carried out special work on jointly training doctoral students with more than ten universities, including Peking University, Tsinghua University, Fudan University, Shanghai Jiao Tong University, Tongji University, the University of Science and Technology of China, and Zhejiang University. The official website also mentions the "North Star" talent program.
【Usage and Limitations】 The official website provides "Join Us" to view the job list, as well as admissions information to learn about the joint doctoral training special program; users need to read and agree to the "User Agreement" and "Privacy Agreement." Regarding technical details such as the specific installation commands, usage steps, license agreements, and platform support for each open-source project, the materials do not currently provide them, and the respective project pages shall prevail. Contact information: comm@pjlab.org.cn; Address: L1 Building, International Media Port, No. 129 Longwen Road, Xuhui District, Shanghai.
