Open-source AI models challenge closed-source dominance amid US-China tech rivalry

China’s AI community unveiled DeepSeek V4 Flash, a model that matches most top Western systems in capability yet is priced dramatically lower and can operate on inexpensive hardware. Restrictions on advanced compute forced Chinese developers to prioritize algorithmic efficiency and open-source releases, turning hardware scarcity into a catalyst for innovation. This development has prompted U.S. figures such as David Sacks, David Friedberg, and Jensen Huang to publicly back open-source AI, while companies like Anthropic adjust their stance toward safety concerns.

Opponents argue that open models could undercut premium services and raise security questions, but proponents note that managed-service APIs keep data within domestic networks. The broader implication is a potential move away from a binary U.S.-China AI race toward a competition between open and closed approaches, with possible collaboration on safety standards.