According to news from this site on September 7, at the 2023 Tencent Global Digital Ecosystem Conference held today, Tencent officially released the Hunyuan large model.
According to reports, the Hunyuan large model parameter scale exceeds 100 billion, and the pre-training corpus exceeds 2 trillion tokens, Tencent More than 50 Tencent businesses and products, including Cloud, Tencent Advertising, Tencent Games, Tencent Financial Technology, Tencent Meetings, Tencent Documents, WeChat Souyisou, and QQ Browser, have been connected to the Tencent Hunyuan large model for testing and have achieved initial results.
can also make a dedicated industry large model based on Hunyuan.
The Hunyuan model claims to be able to identify "traps" and refuses to be "induced" to answer some questions that are difficult or even impossible to answer. Answering questions such as "How to speed better" increases the rejection rate by 20%.
This model can handle ultra-long text in a variety of scenarios. Through positional encoding optimization, long text can be improved. The processing effect and performance of the text. Combined with command follow-up optimization, the output content can better meet the word count requirements.
Hunyuan large model claims to "significantly improve the model's logical thinking ability in scenarios" , strengthening A tendency to break things down and think step-by-step.
Tencent Hunyuan Assistant”. Among them, the icon of “Tencent Hunyuan Assistant” is quite interesting, and the overall design is similar to the traditional Chinese Tai Chi yin and yang pattern.
The functions of "Tencent Hunyuan Assistant" include AI Q&A and AI painting, which can answer various questions and Handle a variety of tasks, such as acquiring knowledge, solving math problems, translating, providing travel guides, work suggestions, etc. Tencent officials have previously demonstrated their ability to write copywriting.
The above is the detailed content of Tencent's Hunyuan large model is officially unveiled: the parameter scale exceeds 100 billion and has been tested on multiple internal products. For more information, please follow other related articles on the PHP Chinese website!