Chinese artificial intelligence startup DeepSeek on Thursday unveiled an upgrade to its flagship V3 model, introducing a feature it says is optimized for compatibility with Chinese-made chips and promising faster processing speeds.
The new release, named DeepSeek-V3.1, underscores Beijing’s push to build a self-reliant semiconductor ecosystem as Washington tightens export restrictions on advanced U.S. technology. Analysts see the move as a clear indication that DeepSeek is aligning its product strategy with China’s broader ambition to reduce dependence on foreign chipmakers.
DeepSeek has already drawn global attention this year by releasing AI models capable of competing with Western rivals such as OpenAI’s ChatGPT. Its products have gained traction in part because they operate at lower costs, making them more accessible to both enterprises and individual users.
Register for Tekedia Mini-MBA edition 19 (Feb 9 – May 2, 2026): big discounts for early bird.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment (next edition begins Jan 24 2026).
The V3.1 upgrade follows a string of recent developments: an R1 model update in May and an earlier enhancement to its V3 model in March. This steady cadence of improvements signals DeepSeek’s determination to maintain pace in the global AI race.
In a WeChat post, the company revealed that DeepSeek-V3.1 incorporates a new UE8M0 FP8 precision format, which has been optimized for “soon-to-be-released next-generation domestic chips.” While DeepSeek did not disclose which manufacturers or chip models will be supported, the development is significant in light of the country’s drive to bolster its own semiconductor sector.
FP8, or 8-bit floating point, is a streamlined data processing format that allows AI models to run faster and more efficiently by consuming less memory compared with traditional methods. The company emphasized that this approach helps reduce computational overhead while improving performance.
Beyond chip compatibility, DeepSeek also introduced a hybrid inference structure to the V3.1 model. This allows the system to operate in both reasoning and non-reasoning modes, depending on the task at hand. Users can seamlessly toggle between these two modes via a “deep thinking” button integrated into DeepSeek’s official app and web platform. Both platforms are now running the upgraded V3.1 model.
In addition to the performance enhancements, DeepSeek announced adjustments to the costs for accessing its API—the platform that enables developers to integrate its AI models into third-party apps and web products. These pricing changes will take effect starting September 6, the company confirmed.
Analysts note that DeepSeek’s strategy mirrors the approach taken by Huawei, which in recent years has been forced to accelerate development of in-house technologies after U.S. sanctions cut off access to critical components. Huawei, once heavily reliant on American chips and software, has since shifted to homegrown semiconductors, proprietary operating systems, and AI frameworks to ensure that its products can function independently of U.S. supply chains.
DeepSeek is thus betting that domestic chipmakers will soon deliver the processing power needed to support cutting-edge AI, allowing the company to scale without being constrained by Washington’s restrictions.
DeepSeek is reinforcing its role as one of China’s most ambitious AI challengers to Western firms by tailoring its AI to domestic chips and expanding usability features. With geopolitical tensions continuing to shape the technology supply chain, the company’s strategy may help accelerate the adoption of homegrown semiconductors while strengthening China’s position in the global AI race.