DeepSeek announced on August 20 that its new V3.1 Base model for Hugging Face was open-sourced. According to the company’s website, the model contains approximately 685 billion parameter values and its context has been increased to 128K.
DeepSeek notified its users earlier that evening that their online model was upgraded to V3.1, with a context length of 128K. This version is available on the official website as well as the app and mini-program. The API call method remains unchanged.
According to the company, there is still no release date confirmed for the highly anticipated DeepSeek R2 Model.[ iThome– in Chinese]
Related
