DeepSeek, a Chinese AI company, has hinted at the imminent arrival of domestically produced, next-generation AI chips. The company mentioned the chips in a social media post about its newly released V3.1 model, stating that the model is specifically designed for upcoming home-grown chips. The V3.1 model incorporates a hybrid inference structure, enabling operation in both reasoning and non-reasoning modes. Users can switch between these modes using a 'deep thinking' button on the company's app and web platform.
The company's focus on domestic chip compatibility suggests an effort to align its AI models with China's developing semiconductor ecosystem, as the country aims to replace US technology amid export restrictions. DeepSeek has gained recognition for its AI models that rival Western models like OpenAI's ChatGPT while offering lower operational costs. The new DeepSeek model is optimised for 'soon-to-be-released next-generation domestic chips' using the UE8M0 FP8 precision format. FP8, or 8-bit floating point, is a data processing format that allows AI models to operate more efficiently, using less memory and running faster than traditional methods. DeepSeek will adjust the costs for using the model's API starting September 6.
