Chinese artificial intelligence startup DeepSeek has unveiled a major upgrade to its flagship V3 model, strengthening its position in the global AI race. The company announced on Thursday that the new version, named DeepSeek-V3.1, features enhanced processing speeds and most notably, compatibility with Chinese-made semiconductor chips.
This focus on domestic chip optimization highlights China’s determination to build a self-reliant technology ecosystem. With ongoing U.S. restrictions on advanced chip exports, Beijing has been encouraging homegrown companies to reduce their dependence on foreign suppliers. DeepSeek’s announcement suggests that its models may soon align with next-generation processors designed by Chinese manufacturers, potentially accelerating the development of the nation’s AI sector.
Domestic Chip Support and Technical Innovation
In a WeChat post, DeepSeek explained that its V3.1 model incorporates the UE8M0 FP8 precision format, specifically designed to work efficiently with upcoming domestic chips. Although the company did not reveal which chipmakers will be supported, the move indicates a forward-looking strategy to ensure long-term viability in China’s fast-changing semiconductor industry.
The FP8 (8-bit floating point) format is particularly important, as it allows AI models to use less memory while achieving faster speeds. This efficiency is critical in large-scale AI applications where computational power and cost-effectiveness often become limiting factors.
DeepSeek details V3.1 and says it surpasses R1 on key benchmarks and is customized to work with next-gen Chinese-made AI chips, after unveiling it on August 19 (Bloomberg)https://t.co/MAvWg7GWMchttps://t.co/TZRS0H13q7— Techmeme (@Techmeme) August 21, 2025
Hybrid Inference Structure
Another key feature of the DeepSeek-V3.1 is its hybrid inference structure, which enables the model to switch between reasoning and non-reasoning modes. Users will be able to activate a special “deep thinking” button within the official DeepSeek app and web platform, giving them greater flexibility in how the model processes complex queries.
This adaptability could set DeepSeek apart from competitors, especially in enterprise and research applications where nuanced problem-solving is essential.
Lower Costs and Competitive Edge
Alongside the technical upgrade, DeepSeek confirmed it will revise API usage costs beginning September 6. This adjustment is expected to make its platform even more accessible for developers and businesses seeking affordable AI integration.
Earlier this year, DeepSeek captured global attention when it launched AI models capable of rivaling Western offerings such as OpenAI’s ChatGPT, but at significantly lower operational costs. The release of the V3.1 update builds on two previous enhancements: the R1 model upgrade in May and an earlier V3 update in March.
DeepSeek’s latest announcement underscores both its technological ambitions and its alignment with national priorities. By fine-tuning its AI models for Chinese-made chips and pushing forward with cost-efficient solutions, DeepSeek is positioning itself not only as a competitor on the global stage but also as a critical player in China’s vision of technological independence.
SEE ALSO: Google Pixel 10 Pro Fold Launched: Price, Specs, Features And AI Innovations
Great Job Omair Pall & the Team @ Mashable India tech Source link for sharing this story.
