DeepSeek Unveils V3.1 AI Model with Domestic Chip Support

DeepSeek has released an upgraded version of its flagship AI model, DeepSeek-V3.1, which is optimized for compatibility with Chinese-made chips. This update is part of the company's strategy to align with China's growing semiconductor ecosystem, as the country seeks to reduce reliance on U.S. technology.

The DeepSeek-V3.1 model introduces a hybrid inference structure, faster processing speeds, and improved agent capabilities. It utilizes the UE8M0 FP8 precision format, which enhances efficiency by using less memory and increasing processing speed. The model is designed to support next-generation domestic chips, although specific chip models or manufacturers have not been disclosed.

This release follows previous updates to DeepSeek's core models, including the R1 model update in May and an earlier V3 enhancement in March. The company also plans to adjust the costs for using the model's API, which allows developers to integrate DeepSeek's AI models into their applications, starting September 6.