The second generation of the InternLM model, InternLM2, includes models at two scales: 7B and 20B. For the convenience of users and researchers, we have open-sourced four versions of each scale of the model, which are:
- internlm2-base-20b: Foundation models with high quality and high adaptation flexibility, which serve as a good starting point for downstream deep adaptations.
- internlm2-20b (**recommended**): Further pretrain with general domain data and domain-enhanced corpus, obtaining state-of-the-art performance in evaluation with good language capability. InternLM2 models are recommended for consideration in most applications.
- internlm2-chat-20b-sft: Intermediate version of InternLM2-Chat that only undergoes supervised fine-tuning (SFT), based on the InternLM2-Base model. We release them to benefit research on alignment.
- internlm2-chat-20b (**recommended**): Further aligned on top of InternLM2-Chat-SFT through online RLHF. InternLM2-Chat exhibits better instruction following, chat experience, and function calling, which is recommended for downstream applications.
The base model of InternLM2 has the following technical features:
- Effective support for ultra-long contexts of up to 200,000 characters: The model nearly perfectly achieves "finding a needle in a haystack" in long inputs of 200,000 characters. It also leads among open-source models in performance on long-text tasks such as LongBench and L-Eval.
- Comprehensive performance enhancement: Compared to the previous generation model, it shows significant improvements in various capabilities, including reasoning, mathematics, and coding.
-`HF` refers to the format used by HuggingFace in [transformers](https://github.com/huggingface/transformers), whereas `Origin` denotes the format adopted by the InternLM team in [InternEvo](https://github.com/InternLM/InternEvo).
We have evaluated InternLM2 on several important benchmarks using the open-source evaluation tool [OpenCompass](https://github.com/open-compass/opencompass). Some of the evaluation results are shown in the table below. You are welcome to visit the [OpenCompass Leaderboard](https://opencompass.org.cn/rank) for more evaluation results.
- The evaluation results were obtained from [OpenCompass](https://github.com/open-compass/opencompass) , and evaluation configuration can be found in the configuration files provided by [OpenCompass](https://github.com/open-compass/opencompass).
- The evaluation data may have numerical differences due to the version iteration of [OpenCompass](https://github.com/open-compass/opencompass), so please refer to the latest evaluation results of [OpenCompass](https://github.com/open-compass/opencompass).
