A new Mozilla report finds that the performance gap between top US closed AI models and the best open-weights models from Chinese developers has narrowed to just 4.4 months, prompting many organisations to switch to cheaper open models for everyday work. This matters because it reshapes how businesses allocate AI spending, with companies increasingly reserving expensive closed models for a narrow band of demanding tasks while defaulting to open alternatives for routine work.

Mozilla's State of Open Source AI report, published on 15 September, found that Moonshot AI's open Kimi K3 model scores just three points behind Anthropic's closed frontier model on a composite intelligence index, despite costing only 30 percent as much. Delivery firm DoorDash reportedly now uses Kimi for routine tasks while reserving the pricier closed model for harder jobs. Mozilla's CTO Raffi Krikorian said closed models retain an edge in expert professional work, intensive retrieval and long-context tasks, and can currently handle jobs around 1.7 times longer than open models can reliably manage—a gap he expects to persist at around four months as both improve.