Chinese AI startup DeepSeek has released an upgrade to its flagship V3 model, adding a feature the company says can optimize it for Chinese-made chips and offering faster processing, as detailed in a WeChat post on Thursday. According to Reuters, the update underscores a focus on compatibility with China’s emerging semiconductor ecosystem.
V3.1 focuses on domestic chip readiness
DeepSeek said the DeepSeek-V3.1 model supports a UE8M0 FP8 precision format tailored for “soon-to-be-released next-generation domestic chips.” The company did not name specific chips or manufacturers. FP8, or 8-bit floating point, is described as a data processing format that can allow AI models to operate more efficiently by using less memory and running faster than traditional methods.
The emphasis on domestic chip support comes as Beijing promotes broader replacement of U.S. technology. Reuters noted this push follows Washington’s export restrictions and suggests DeepSeek’s models are being positioned to align with China’s semiconductor development.
Hybrid inference and product rollouts
DeepSeek-V3.1 introduces a hybrid inference structure that the company says enables operation in both reasoning and non-reasoning modes. Users can switch between these modes via a “deep thinking” button on DeepSeek’s official app and web platform, both of which now run V3.1.
Model updates and API pricing changes
The V3.1 upgrade follows two other recent enhancements to DeepSeek’s core lineup: an R1 model update in May and an earlier V3 enhancement in March. The company also stated it will adjust costs for using the model’s API starting September 6, affecting developers who integrate its AI models into other apps and web products.
Earlier this year, DeepSeek drew attention by releasing AI models that Reuters said compete with Western offerings like OpenAI’s ChatGPT while providing lower operational costs. With V3.1, DeepSeek highlights performance and chip compatibility features while expanding user controls through its platforms.
Reuters reported that the company’s announcements were made via WeChat posts and that no additional chip details were provided. The move situates DeepSeek’s latest upgrade within ongoing efforts to bolster domestic hardware support amid industry shifts and policy dynamics.
