Kunlun Wanwei released hundreds of billions of large language model "Tiangong", inviting testing to start.

CTOnews.com April 17 news, Kunlun Wanwei officially released the 100 billion-level large language model "Tiangong", and announced that the invitation test will be launched from now on, and interested users can click here to enter the official.

CTOnews.com learned that "Tiangong" is jointly developed by Kunlun Wanwei and AI team Singularity Zhiyuan, and interacts with users through natural language. AI generation ability can meet the needs of copywriting, knowledge Q & A, code programming, logical deduction, mathematical calculation and other needs, covering science, technology, culture, art, history and other fields.

Official introduction, "Tiangong" has been able to support more than 10,000 words of text dialogue, achieve more than 20 rounds of user interaction. "Tiangong" calculation is based on one of the largest GPU clusters in China, using a double 100 billion model-100 billion pre-training base model and 100 billion RLHF model, through the Monte Carlo search tree algorithm to provide a more humane interactive experience.