DeepSeek claims the V4 model can stand shoulder to shoulder with top American systems from Google, OpenAI, and Anthropic.
DeepSeek claims the V4 model can stand shoulder to shoulder with top American systems from Google, OpenAI, and Anthropic.
The Chinese AI firm DeepSeek unveiled a preview of its highly awaited next-gen AI model V4 this past Friday, stating that the open-source model can rival leading closed-source systems from American competitors such as Anthropic, Google, and OpenAI.
DeepSeek asserts that V4 signifies a substantial upgrade over its previous models, particularly in coding, a feature that has become crucial for AI agents and has contributed to the popularity of tools like ChatGPT Codex and Claude Code. The launch is also a significant step for China’s chip industry, as DeepSeek specifically emphasizes compatibility with domestic Huawei technology.
This release follows a year after DeepSeek shook the US AI sector with R1, a model it claimed was trained at a fraction of the expense of top American systems. DeepSeek has not revealed the training costs for V4 or the hardware utilized. US officials have alleged that the company has utilized banned Nvidia chips and Anthropic asserts that DeepSeek misappropriated Claude to enhance its own offerings.
