DeepSeek Boosts AI Model with Chip-Friendly Upgrade
Chinese artificial intelligence startup DeepSeek has unveiled an upgraded version of its flagship model, DeepSeek-V3.1, with faster performance and improved compatibility for domestic chips. The announcement was made on Thursday through the company’s WeChat post.
The new model is designed to align with China’s growing semiconductor sector, which has gained momentum as Beijing continues efforts to reduce dependence on U.S. technology amid Washington’s export restrictions. Analysts suggested the update underlines China’s determination to strengthen its local technology ecosystem.
DeepSeek has already gained attention for producing AI models that rival Western competitors such as OpenAI’s ChatGPT, while keeping operational costs lower. The release of V3.1 follows earlier enhancements this year, including an R1 model in May and a previous V3 update in March.
Currency Converter
Read More: UK Monarch Offers Support to Flood-Hit Pakistan
According to the company, the model introduces support for FP8 (8-bit floating point) precision format, which is expected to boost efficiency by cutting memory use and speeding up processes. While DeepSeek did not name the specific chipmakers, it indicated the model was optimized for “next-generation domestic chips.”
The V3.1 also includes a hybrid inference system that allows users to switch between reasoning and non-reasoning tasks. A “deep thinking” button has been added to its application and web platform, offering users flexibility in how the model processes information.
DeepSeek confirmed that API usage charges would be revised from 6 September, enabling developers to integrate the model into apps and services with updated pricing.
