tech
China's DeepSeek releases preview of long-awaited V4 model as AI race intensifies
Chinese AI startup DeepSeek has released a preview version of its long awaited V4 large language model.
TL;DR
- DeepSeek released a preview version of its V4 large language model, which is open-source.
- V4 offers lower inference costs and strong performance, particularly in agent-based tasks.
- The model is available in 'pro' and 'flash' versions and is optimized for tools like Anthropic's Claude Code.
- DeepSeek's previous R1 model had raised concerns about the US lead in AI and Big Tech's spending.
- The V4 release intensifies competition among Chinese AI players like Alibaba and ByteDance.
- Huawei confirmed its AI cluster can support DeepSeek's V4 model, raising questions about chip usage, especially given US export controls on Nvidia chips.
- The ability to run natively on local chips could boost China's AI sovereignty.