Exclusive: Open Chinese models close gap with Silicon Valley’s frontier AI models

A new report from Mozilla highlights a significant shift in the artificial intelligence landscape, revealing that the performance gap between top-tier US frontier models and open-weights Chinese models has narrowed to just 4.4 months. The study, which references the Artificial Analysis Intelligence Index, notes that Moonshot AI’s Kimi K3 model now performs within three points of Anthropic’s Fable 5 while costing only 30 percent of the price. Mozilla CTO Raffi Krikorian suggests that organizations should prioritize open models for most routine tasks, reserving expensive closed-source frontier models only for specialized, high-intensity workloads such as expert professional analysis and long-context retrieval. This trend indicates that the premium paid for proprietary models is becoming increasingly workload-specific rather than a necessity for entire organizations, marking a pivotal moment for the adoption of open-source AI in the global enterprise sector.