Baichuan2–192K LLM is released

Baichuan2–192K LLM is released

The context window length is as high as 192K, which is currently the longest context window in the world and can process approximately 350,000 Chinese characters at a time.

Official announcement: Baichuan2–192K not only surpasses Claude2 in terms of context window length, but also leads Claude2 in terms of long window text generation quality, long context understanding, long text Q&A, and summarization.

7 of the 10 long text evaluations achieved SOTA, leading Claude2 in all aspects

Baichuan2–192K performed well on 10 evaluation sets of Chinese and English long text question answering and summarization, including Dureader, NarrativeQA, LSHT, and TriviaQA. Seven of them achieved SOTA, significantly surpassing other long window models.

In addition, LongEval’s evaluation results show that Baichuan2–192K can still maintain very strong performance after the window length exceeds 100K, while the performance of other open source or commercial models has almost plummeted after the window length increases. Claude2 is no exception. The overall effect drops very seriously after the window length exceeds 80K.

Baichuan2–192K has officially started internal testing and has been implemented in many real-life scenarios such as legal and media.

Baichuan2–192K has officially started internal testing and is open to Baichuan Intelligence’s core partners through API calls. It has reached cooperation with financial media, law firms and other institutions to apply Baichuan2–192K’s world-leading long context capabilities. In specific scenarios such as media, finance, and law, it will be fully opened in the near future.

Detailed: mp.weixin.qq.com/s/lAJh6qGG27u_

More AI News