Baichuan Intelligent announced the opening of Baichuan2-Turbo series API, allowing users to create their own knowledge base.

CTOnews.com, December 19, at the Baichuan2 Turbo series API conference held today, Baichuan Intelligent announced the opening of the search-enhanced Baichuan2-Turbo series API, two models "Baichuan2-Turbo-192K" and "Baichuan2-Turbo", adding the ability of "search enhanced knowledge base" on the basis of "supporting 192 K context window".

Wang Xiaochuan, the ▲ source Baichuan intelligent official website, claims that Baichuan Intelligence, through a long window + search enhancement, increases the text size that the large model can obtain to 50 million tokens "equivalent to 100 million Chinese characters" on the basis of the 192 K long context window. the model can first retrieve the most relevant content in a large number of documents according to the user's prompt words, and then put these documents into the long window together with the prompt words. Thus the reasoning cost and time cost can be saved.

CTOnews.com noted that Baichuan Intelligence claimed that from now on, API users can upload text materials to create an "exclusive knowledge base". At present, the official website large model already supports PDF, Word and other text uploads, as well as URL URL input. Users can experience the improved model through the entry of the official website, and interested partners can click here to visit.

Related reading: "Baichuan Intelligence launched the Baichuan2-19K model, and the context window is about 14 times longer than GPT-4."