On July 3, 2024, the Shanghai AI Lab, in collaboration with SenseTime Technology, the Chinese University of Hong Kong, and Fudan University, officially released the next-generation large language model, InternLM2.5.
The InternLM2.5-7B model has been open-sourced, and models of other scales will be released in sequence. The Shanghai AI Lab promises to continue to provide free commercial licenses and support community innovation with high-quality open-source models.
The model has significant improvements in the following aspects:
Inference Capability: The inference capability of InternLM2.5 has been significantly enhanced, surpassing the Llama3-70B model in some dimensions. Especially on the MATH evaluation set, it has achieved a 100% performance improvement, with an accuracy rate of 60%, comparable to the GPT-4Turbo1106 version.
Support for Long Texts: The model supports contexts of up to 1 million tokens, which can handle about 1.2 million Chinese characters of long texts. By increasing the length of context and synthesizing data, it optimizes the understanding of long documents and intelligent agent interactions.
Autonomous Planning and Tool Invocation: With the ability to search and integrate information from hundreds of web pages, the model simulates the thinking process of humans through the MindSearch multi-agent framework, effectively integrating network information.
Github Link: https://github.com/InternLM/InternLM
Model Link: https://www.modelscope.cn/models/Shanghai_AI_Laboratory/internlm2_5-7b-chat
InternLM2.5 Homepage: https://internlm.intern-ai.org.cn/