OpenAI CEO Sam Altman said Thursday that Chinese open-source artificial intelligence models are getting very good during an interview with CNBC [1, 2].

The admission highlights the intensifying global competition in the AI sector, as open-source developments from China challenge the dominance of U.S.-based proprietary systems.

Speaking during CNBC's "Squawk on the Street," Altman said the rapid progress of international competitors is notable [1, 2]. Despite the rise of these models, he expressed confidence in the trajectory of his company's technology. "We will continue to have the best models in the world," Altman said [1, 2].

Altman also discussed specific performance gains within OpenAI's latest releases. He said that the new GPT-5.6 Sol model is 54% more token-efficient on agentic coding tasks [3]. This efficiency represents a technical leap in how the AI handles complex programming workflows, a key area of competition between the U.S. and China.

The conversation comes as OpenAI navigates headwinds related to compute and the pressure to maintain a significant lead over both open-source and corporate rivals [4]. While open-source models allow for wider accessibility and customization, OpenAI relies on its closed-source superiority to maintain its market position [1, 2].

Altman's comments suggest a strategy of acknowledging the narrowing gap in capabilities while focusing on specialized efficiency and raw power to stay ahead [1, 3].

“"Chinese open-source models are getting very good."”

The acknowledgment of Chinese AI progress signals a shift from a period of undisputed U.S. dominance to a more fragmented competitive landscape. By emphasizing token efficiency in the GPT-5.6 Sol model, OpenAI is attempting to pivot the definition of 'best' from general capability to operational efficiency and agentic performance, which are critical for enterprise adoption.