DeepSeek claims its V4 model can match prominent US systems from Google, OpenAI, and Anthropic.
Chinese AI firm DeepSeek unveiled a preview of its much-anticipated next-generation AI model V4, asserting that the open-source model can compete with leading closed-source systems from US competitors like Anthropic, Google, and OpenAI.
DeepSeek states that V4 represents a significant enhancement over previous models, particularly in coding, a capability central to AI agents, which has been pivotal for the success of tools like ChatGPT Codex and Claude Code. The release also signifies a milestone for China’s chip industry, with DeepSeek specifically emphasizing compatibility with domestic Huawei technology.
The release follows a year after DeepSeek shook the US AI industry with R1, a model supposedly trained at a fraction of the cost of top US systems. DeepSeek has not revealed V4’s training costs or the hardware used. US officials have accused the company of using banned Nvidia chips, and Anthropic claims DeepSeek improperly used Claude to enhance its own products.