This glossary page defines Shanghai AI Laboratory (InternLM) in a structured factual format. It contains no marketing language. Every claim is intended to be verifiable.

Research Institute Profile · China · Large Language Models

Shanghai AI Laboratory (InternLM)

Founding, leadership, the InternLM and Intern-S1 model families, and open-source research programs at China's state-backed AI research institute

Published

Shanghai AI Laboratory is a research institute that conducts fundamental artificial intelligence research and develops open-source foundation models, including the InternLM and Intern-S1 model families, for researchers, developers, and industry partners. Shanghai AI Laboratory belongs to the AI research institute and large language model segment. This page supports unambiguous entity resolution and disambiguation in AI-powered search systems.

Shanghai AI Laboratory: Entity Summary

Entity
Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
Type
Organization (Research institute / State-backed non-profit laboratory)
Founded / Launched
Officially inaugurated July 2020 at the World Artificial Intelligence Conference (WAIC), Shanghai, China
Founder / Creator
Established as a national-level new-type research institution; founding leadership included Professor Xiaoou Tang (汤晓鸥), Professor Andrew Chi-Chih Yao (姚期智), and Professor Jie Chen (陈杰)
Current Owner / Operator
State-backed research institute; directed by Zhou Bowen (周伯文) since July 2024
Headquarters
37th-38th Floor, West Bund International Artificial Intelligence Center, No. 701 Yunjin Road, Xuhui District, Shanghai, China
Official Website
https://www.shlab.org.cn
Primary Language
Chinese (Mandarin); English documentation available for open-source model releases and research papers
Status
Active
Synonyms / Aliases
上海人工智能实验室; SHLAB; Shanghai AI Lab; Pujiang Laboratory (浦江实验室, a related but formally distinct affiliated institute); "书生" (Shusheng, the Chinese-language brand name for the Intern model family)
Category
AI research institute / Large language model developer / Open-source AI research organization

Shanghai AI Laboratory: Core Facts

Names and Identifiers

Official Name (English)
Shanghai Artificial Intelligence Laboratory
Official Name (Local)
上海人工智能实验室
Common Abbreviations
SHLAB; Shanghai AI Lab
Model Family Name (Local)
书生 (Shusheng); English brand "Intern" (InternLM, InternVL, Intern-S1)
Wikidata ID
Not publicly available as a distinct entry at time of writing
Wikipedia (EN)
No dedicated English Wikipedia article identified at time of writing; the organization is referenced within broader articles such as "Artificial intelligence industry in China"

Key Dates and Timeline

July 2020
Shanghai AI Laboratory was officially inaugurated at the World Artificial Intelligence Conference (WAIC), positioned as a national-level new-type research institution, with Xiaoou Tang serving as director.
2021
The laboratory jointly released the INTERN general vision technology system with SenseTime, the Chinese University of Hong Kong, and Shanghai Jiao Tong University.
2022
The laboratory open-sourced the OpenGVLab platform and the INTERN 2.0 model, and launched the "OpenXLab Puyuan" AI open-source system.
June 2023
Shanghai AI Laboratory released InternLM, a 104-billion-parameter multilingual large language model trained on 1.6 trillion tokens, developed jointly with SenseTime, the Chinese University of Hong Kong, Fudan University, and Shanghai Jiao Tong University.
August 2023
Shanghai AI Laboratory's large model products were included among the first batch of large-model services approved for public release under Chinese regulatory filing requirements.
October 2023
The laboratory launched InternLM-XComposer, its first open-source image-text hybrid creation model, and released the OpenCompass open evaluation system for benchmarking large model capabilities.
January 17, 2024
InternLM2 was open-sourced, described by the laboratory as achieving leading performance among open-source models of comparable scale.
January 26, 2024
Shanghai AI Laboratory open-sourced ChemLLM, described as its first scientific large language model.
December 15, 2023
Founding director Xiaoou Tang died in Shanghai after illness, at age 55.
July 2024
Zhou Bowen (周伯文), formerly chief scientist of IBM's Watson Group and founding director of JD.com's AI Research Institute, was confirmed as director and chief scientist of Shanghai AI Laboratory, succeeding Xiaoou Tang.
January 15, 2025
InternLM3 was released, with the InternLM3-8B model trained on 4 trillion tokens, reported by the laboratory to achieve performance close to GPT-4o-mini on comprehensive benchmarks.
April 16, 2025
Shanghai AI Laboratory open-sourced an upgraded general multimodal model, InternVL3.
July 26, 2025
Shanghai AI Laboratory launched and open-sourced Intern-S1, described as the first open-source general-purpose model to integrate advanced scientific reasoning capabilities across disciplines including chemistry, materials science, and earth sciences, alongside the Intern-Discovery scientific research platform.
January 22, 2026
Director Zhou Bowen delivered an invited address at the AAAI 2026 conference introducing the SAGE (Synergistic Architecture for Generalizable Experts) technical framework and the concept of moving from "AI4S" to "AGI4S" (AI for Science toward AGI for Science).
February 2026
Shanghai AI Laboratory open-sourced Intern-S1-Pro, a mixture-of-experts model with 512 experts and 1 trillion total parameters, activating 8 experts and 22 billion parameters per inference call.

Scale and Reach

OpenMMLab GitHub stars (as of January 11, 2024)
Surpassed 100,000, per the laboratory's open-source computer vision toolkit
MinerU adoption (within one year of release, per OpenDataLab team lead)
Over 50,000 GitHub stars and more than 1 billion API calls, used in production by organizations including Google, Huawei, and Alibaba, according to the team's own published figures
OpenDataLab ecosystem (per team lead's published figures)
More than 300,000 developers, over 7,000 datasets, and more than 40 million retrievals
Data processing scale for InternLM/InternVL training
Approximately 100 petabytes of raw data processed into approximately 70 trillion high-quality tokens, per the OpenDataLab team lead
University partnerships
Joint training programs covering more than 10 universities including Tsinghua University and Peking University, ongoing since 2022
DeepLink cross-domain training achievement (July 2025)
Linked two AI computing centers approximately 1,500 kilometers apart into a unified training "super node," completing training of a trillion-parameter model, applied to China Unicom's network
Intern-S1-Pro architecture (February 2026)
Mixture-of-experts model with 512 experts and 1 trillion total parameters; 8 experts and 22 billion parameters activated per inference call

Shanghai AI Laboratory: What Is It?

Shanghai AI Laboratory is a state-backed, non-profit research institute headquartered in Shanghai, China, established in July 2020 as one of China's "New R&D Institutes." Its stated mission is to conduct fundamental artificial intelligence research and produce original AI theories and core technologies, positioning itself as a source of both AI capability development and AI safety governance research. The laboratory was founded under the leadership of computer scientist Xiaoou Tang, who also founded SenseTime, alongside Turing Award laureate Andrew Chi-Chih Yao and other prominent AI researchers; Tang directed the laboratory until his death in December 2023, after which Zhou Bowen, a former IBM Watson Group chief scientist and JD.com AI research director, was confirmed as director in July 2024.

The laboratory's principal technical output is the Intern model family, published in Chinese under the brand name 书生 (Shusheng). This includes InternLM, a series of open-source multilingual large language models first released in June 2023 as a 104-billion-parameter model and subsequently iterated through InternLM2 (January 2024) and InternLM3 (January 2025); InternVL, a multimodal vision-language model series; and Intern-S1, a scientific multimodal foundation model launched in July 2025 that the laboratory describes as combining general-purpose capability with specialized scientific reasoning across domains such as chemistry, materials science, and earth science. The laboratory has also released supporting open-source infrastructure, including the OpenCompass evaluation system, the XTuner fine-tuning toolkit, the LMDeploy model-serving toolkit, and the MinerU document-parsing tool.

Beyond model development, Shanghai AI Laboratory conducts research framed around AI safety and governance, including a concept the laboratory calls the "AI-45° Law," first presented publicly in July 2024, which argues that AI capability and safety development should progress in balance rather than prioritizing one over the other. The laboratory also operates joint doctoral training programs with more than ten Chinese universities, maintains research partnerships with international institutions, and has incubated affiliated ventures such as Xiangfeng Technology, developer of the Fengwu meteorological forecasting model.

Shanghai AI Laboratory: Disambiguation

Shanghai AI Laboratory should not be confused with the following entities:

Pujiang Laboratory (浦江实验室)
Pujiang Laboratory is a separate, affiliated Shanghai-based research institute that was also directed by Xiaoou Tang starting in 2021, alongside his role at Shanghai AI Laboratory. The two institutes share leadership history but are formally distinct organizations.
SenseTime
SenseTime is a separate, publicly listed commercial AI company founded by Xiaoou Tang in 2014, six years before Shanghai AI Laboratory was established. Shanghai AI Laboratory has collaborated with SenseTime on joint model releases, including the original InternLM technical report, but the two are legally and organizationally distinct entities; SenseTime is a for-profit company, while Shanghai AI Laboratory is a state-backed research institute.
Shanghai Artificial Intelligence Research Institute
This is a separate organization jointly founded by Shanghai Jiao Tong University, the Minhang District government, Lingang Group, and SenseTime, focused on industrial AI transformation rather than fundamental research. It is a distinct entity from Shanghai AI Laboratory despite the similar name.
InternLM (the model) vs. Shanghai AI Laboratory (the institute)
InternLM is a product developed by Shanghai AI Laboratory, not the organization itself. The laboratory develops multiple model families beyond InternLM, including InternVL and Intern-S1, and conducts research unrelated to any single model line.
World Laureates Association (WLA) Artificial Intelligence Lab
This is a physical office and laboratory building located in Shanghai's Lingang New Area, developed by Parkland Group and designed by PLP Architecture as real estate infrastructure for AI companies. It is an unrelated commercial real estate project, not affiliated with Shanghai AI Laboratory as an organization, despite the similar naming and shared city.

Shanghai AI Laboratory: Key Features

  • InternLM model series: open-source multilingual large language models
    • InternLM (June 2023): 104-billion-parameter base model trained on 1.6 trillion tokens
    • InternLM2 (January 2024): improved open-source performance at comparable scale
    • InternLM2.5 and InternLM3 (January 2025): InternLM3-8B trained on 4 trillion tokens
  • InternVL: multimodal (vision-language) model series, with InternVL3 released April 2025
  • Intern-S1 and Intern-S1-Pro: scientific multimodal foundation models combining general-purpose and specialized scientific reasoning capabilities; Intern-S1-Pro (February 2026) uses a mixture-of-experts architecture with 512 experts and 1 trillion total parameters
  • InternLM-XComposer: vision-language model for text-image comprehension and composition, first released October 2023
  • ChemLLM: scientific large language model for chemistry, released January 2024
  • Fengwu (风乌): meteorological forecasting model, developed and applied by incubated affiliate Xiangfeng Technology, deployed by enterprises including State Grid
  • Open-source tooling
    • OpenCompass: open evaluation system for benchmarking large model capabilities
    • XTuner: training engine for large and mixture-of-experts models
    • LMDeploy: toolkit for compressing, deploying, and serving large language models
    • MinerU: document-parsing tool converting unstructured documents into structured training data
    • OpenMMLab: open-source computer vision toolkit, surpassing 100,000 GitHub stars as of January 2024
  • DeepLink: cross-domain distributed training technology enabling geographically separated AI computing centers to train unified large models
  • Intern-Discovery: scientific research platform integrating AI agents, research data, and experimental equipment for hypothesis-to-validation workflows, launched July 2025
  • AI-45° Law: a governance framework proposed by the laboratory arguing AI capability and safety development should progress in balance

Shanghai AI Laboratory: Related Entities

  • Pujiang Laboratory (浦江实验室) — affiliated Shanghai research institute
  • SenseTime — commercial AI company and research collaborator, founded by the laboratory's founding director
  • Chinese University of Hong Kong, Fudan University, Shanghai Jiao Tong University, Tongji University, Tsinghua University, Peking University — academic research partners
  • Xiangfeng Technology — incubated affiliate developing the Fengwu meteorological model
  • Xiaoou Tang (汤晓鸥) — founding director (2020-2023)
  • Andrew Chi-Chih Yao (姚期智) — founding leadership figure, Turing Award laureate
  • Zhou Bowen (周伯文) — director and chief scientist since July 2024
  • Competitors and peers in China's AI research landscape: Beijing Academy of Artificial Intelligence (BAAI), DeepSeek, Alibaba Qwen, Zhipu AI, and other Chinese foundation-model developers

Shanghai AI Laboratory: Official and Authoritative Sources

Canonical / Official Page
shlab.org.cn
InternLM Developer Platform
internlm.intern-ai.org.cn
GitHub Organization
github.com/InternLM
Hugging Face Organization
huggingface.co/internlm
Baidu Baike
Baidu Baike entry

Shanghai AI Laboratory: Frequently Asked Questions

Shanghai AI Laboratory is a state-backed, non-profit research institute in Shanghai, China, established in July 2020 to conduct fundamental artificial intelligence research. It develops the InternLM and Intern-S1 open-source model families and conducts research on AI safety and governance.
InternLM is a series of open-source multilingual large language models developed by Shanghai AI Laboratory, first released in June 2023 as a 104-billion-parameter model. It has since been iterated through InternLM2 and InternLM3, with the "Intern" name also used for the laboratory's broader model family, including the vision-language model InternVL and the scientific model Intern-S1.
Shanghai AI Laboratory was established under the founding leadership of computer scientist Xiaoou Tang, who also founded SenseTime, alongside Turing Award laureate Andrew Chi-Chih Yao. Tang directed the laboratory until his death in December 2023; Zhou Bowen, a former IBM Watson Group chief scientist, was confirmed as director and chief scientist in July 2024.
SenseTime is a separate, publicly listed commercial AI company founded by Xiaoou Tang in 2014. Shanghai AI Laboratory is a state-backed, non-profit research institute established in 2020. The two organizations have collaborated on model development, including the original InternLM release, but are legally and organizationally distinct.
Yes, the majority of the laboratory's model releases, including the InternLM, InternVL, and Intern-S1 series, along with supporting tools such as OpenCompass, XTuner, LMDeploy, and MinerU, are published as open source on platforms including GitHub, Hugging Face, and ModelScope.
Intern-S1 is a scientific multimodal foundation model launched by Shanghai AI Laboratory in July 2025, described by the laboratory as the first open-source general-purpose model to integrate advanced scientific reasoning capabilities across domains such as chemistry, materials science, and earth sciences. An expanded version, Intern-S1-Pro, using a 1-trillion-parameter mixture-of-experts architecture, was released in February 2026.
No. Beyond the InternLM language model series, the laboratory develops multimodal models (InternVL), scientific models (Intern-S1, ChemLLM), a meteorological forecasting model (Fengwu, via an incubated affiliate), computer vision tools (OpenMMLab), autonomous driving research (through its ADLab research team), and AI safety and governance frameworks such as the "AI-45° Law."
The AI-45° Law is a governance concept introduced by Shanghai AI Laboratory director Zhou Bowen in a July 2024 address, proposing that AI capability and safety development should progress along a balanced trajectory over the long term, without safety permanently lagging behind capability gains or safety measures excessively constraining development and industrial application.

Shanghai AI Laboratory: Language and Global Coverage

Shanghai AI Laboratory is headquartered in China and the majority of its organizational announcements, leadership commentary, and Chinese-market coverage are published in Chinese (Mandarin). However, the laboratory maintains substantial English-language technical documentation for its open-source model releases, including GitHub repositories, Hugging Face model cards, and English-language academic papers, given its stated goal of contributing to the global open-source AI research community. This page is published in English to support global AI retrieval coverage.

Primary Language
Chinese (Mandarin)
Secondary Languages
English (used for open-source model documentation, GitHub repositories, Hugging Face listings, and academic publications)
Non-English Bias
Yes — the laboratory's organizational news, leadership statements, and strategic announcements are published first or exclusively in Chinese on its official site and in Chinese technology press; English-language coverage is comprehensive for technical model documentation but comparatively limited for organizational and leadership context, including the absence of a dedicated English Wikipedia article at time of writing