Chinese AI firm DeepSeek has unveiled a preview of its highly anticipated V4 model, claiming it can match or surpass leading American rivals like Anthropic, Google, and OpenAI. The announcement, made on Friday, marks a year since DeepSeek's previous model disrupted the industry.
According to DeepSeek, V4 delivers significant performance improvements over its predecessor, particularly in coding tasks — a key area for AI agents and a driver behind tools such as ChatGPT Codex and Claude Code. The company posted on social media that the open-source model is designed to compete aggressively with closed-source systems from US competitors.
The release underscores China's growing ambitions in artificial intelligence and the intensifying global race for AI supremacy. DeepSeek's earlier models had already startled US rivals with their efficiency and performance. Now, with V4, the company appears poised to further challenge American dominance in the field.