Chinese AI startup DeepSeek has announced the release of its upgraded V3.1 model, featuring domestic chip compatibility, faster processing speeds, and enhanced efficiency. The update underscores China’s push to build a self-reliant semiconductor ecosystem amid U.S. export restrictions, while positioning DeepSeek as a key competitor to Western AI platforms like OpenAI’s ChatGPT.
DeepSeek’s Push for Domestic Chip Optimisation
The highlight of the new DeepSeek-V3.1 upgrade is its support for Chinese-made next-generation chips. According to the company’s statement on WeChat, the model is optimised with the UE8M0 FP8 precision format, designed to deliver:
- Faster inference speeds
- Reduced memory usage
- Higher efficiency compared to traditional data formats
Although DeepSeek has not revealed which chip models or manufacturers will be supported, the move aligns with Beijing’s strategy to reduce reliance on U.S. technology.
Key Features of the DeepSeek-V3.1 Model
The upgraded model introduces several new capabilities:
- Hybrid inference structure: Allows the AI to operate in both reasoning and non-reasoning modes.
- “Deep Thinking” toggle: Users can switch between different reasoning levels on DeepSeek’s official app and web platform.
- FP8 efficiency: The 8-bit floating point (FP8) system reduces computing costs while improving speed and performance.
This innovation positions DeepSeek’s technology as both cost-effective and scalable for developers and enterprises.
Competitive Edge in the Global AI Race
DeepSeek gained global attention earlier this year for offering AI models with lower operational costs compared to Western alternatives. With the release of V3.1, the company strengthens its ability to compete in the rapidly expanding AI development market, particularly in regions looking for non-U.S. technology alternatives.
The company’s recent rollout follows earlier updates:
- DeepSeek R1 model (May 2025)
- V3 enhancement (March 2025)
API Pricing Adjustment
Alongside the release, DeepSeek announced that it will adjust API costs starting September 6, 2025. This move aims to make the upgraded DeepSeek-V3.1 model more accessible for developers integrating AI into apps, software, and web platforms.
