Chinese company claims to surpass OpenAI in long text processing

Baichuan2-192k is the latest large language model (LLM) from Baichuan, the company behind China's popular search engine Sogou. Founder Wang Xiaochuan said the new LLM, based on a "Context Window," can process about 350,000 Chinese characters, making it the world's most powerful model for processing long text statements.

The context window is the combination of input and output text that a model can process during a conversation with a user. According to the WeChat post, Baichuan2-192k has 14 times the processing power of GPT-4, the large language model in OpenAI's ChatGPT.

LLM has the world's previous largest contextual window size, which was announced in July by Amazon-backed Anthropic's Claude 2. The model can handle contextual window data of up to 75,000 English words, which is equivalent to hundreds of pages of a document or a book. If Baichuan's claim is correct, Baichuan2-192k is nearly five times more powerful than Claude 2.

Baichuan claims that its model surpasses Claude 2 in terms of response quality and ability to understand and summarize long texts. This claim is based on the test results of LongEval, a project initiated by the University of California, Berkeley and other US institutions to evaluate the processing level of a particular LLM model.

According to Xiaochuan, Baichuan2-192k is useful for businesses that need to process and create long documents on a daily basis, such as the legal, media, and financial industries. The company is testing the model internally with some partners.

However, according to research by scholars from Stanford University and UC Berkeley, processing more information does not necessarily make an AI model better. Before Baichuan, several Chinese LLMs also claimed to have surpassed ChatGPT. On October 31, Alibaba said that Tongyi Qianwen, an AI model trained with hundreds of billions of parameters, had surpassed OpenAI's GPT-3.5 and Meta's Llama2, and "significantly narrowed the gap" with GPT-4. Meanwhile, Zhipu AI, a startup backed by Alibaba and Tencent, last week launched ChatGLM3 with several improvements, including faster inference speed, lower training costs, and the addition of a coding assistant.