Mozilla Report: U.S.-China Open-Source AI Gap Narrows to 4.4 Months
Mozilla released its "State of Open Source AI" report, revealing that the capability gap between U.S. and Chinese open-weights AI models has narrowed to approximately 4.4 months. The findings track benchmark progress across leading open-source large language models from both countries. This rapid narrowing highlights China's strong momentum in open-source AI development despite U.S. hardware export restrictions. It suggests that control over global AI capability is shifting toward a more competitive, multipolar landscape in the open-source ecosystem. The study specifically evaluates open-weights LLMs over time across standard evaluation benchmarks. It shows Chinese research entities are releasing high-performing models just months after top U.S. labs establish new state-of-the-art benchmarks.
## BACKGROUND
Open-weights AI models are AI systems whose trained parameter weights are made publicly available for developers to download, modify, and run locally. Historically, leading U.S. AI labs held a multi-year lead in frontier model development. However, strong investments by Chinese tech firms and academic institutions into open models have rapidly closed the technical distance.