The Chinese startup DeepSeek presented the updated version of its AI Modeli-V3.1, which received an expanded context for processing long dialogs.
This will allow the system to better keep the details in memory and conduct more coherent conversations even with a large amount of data.
The announcement took place in the official group of the company in WeChat. There is no detailed documentation yet and the model has not been posted on platforms such as Huging Face.
However, the fact of the release of V3.1 shows the rapid rate of development of Deepseek and the desire to compete with Western developers.
The company from Hangzhou has already attracted attention thanks to the R1 model, which in a number of tests went around the Western analogues.
Now users are waiting for the release of R2. True, its launch is postponed – according to media reports, it is a matter in the desire of the CEO of Startup, Liang Wenfen, to bring the model to the ideal and eliminate technical errors.
