Shanghai AI Laboratory releases InternLM

Shanghai AI Laboratory, working jointly with SenseTime and in collaboration with the Chinese University of Hong Kong, Fudan University and Shanghai Jiao Tong University, open-sourced InternLM, a large language model released in two sizes — a 7-billion and a 104-billion-parameter version — with the larger model pre-trained on 1.6 trillion tokens of Chinese and English text in a multi-phase process, then fine-tuned to align with human preferences. Code was published under the Apache 2.0 licence, with weights free for academic research and, in most cases, commercial use.

InternLM was pitched as a general-purpose foundation model with support for long-context understanding and tool use, and the project positioned itself explicitly against contemporary Western open releases such as LLaMA and Falcon, aiming for competitive performance on both English- and Chinese-language benchmarks. Later InternLM releases — InternLM2 in 2024, InternLM2.5, and the multimodal InternVL and Intern-S1 lines — extended the same lab’s open-release strategy, with training-efficiency claims and benchmark scores improving across successive versions.

The release is notable chiefly as an early marker that China’s open-model ecosystem was not confined to a single national champion: Shanghai AI Laboratory, an academically-affiliated, government-linked research institute rather than a commercial lab, was shipping openly licensed, competitively benchmarked models on the same cadence as Meta, TII and the Together AI-led RedPajama effort, at a point when most Western commentary on Chinese AI still focused on Baidu and Alibaba.