A new Mozilla report reveals the performance gap between US frontier AI models and the best open-weights models from Chinese companies has closed to just 4.4 months. This shift is driving companies to adopt significantly cheaper open models for routine work, reserving expensive closed models only for specialized tasks. The report highlights that leading open model Kimi K3 from Moonshot AI trails Anthropic's Fable 5 by only three points on the Artificial Analysis Intelligence Index, despite costing just 30 percent as much.

Raffi Krikorian, Mozilla's chief technology officer, explained that the decision to pay for closed models should be workload-specific rather than organization-specific. Organizations are increasingly using open models as the default for the majority of their workloads. For example, delivery company DoorDash has been using Kimi for routine tasks while reserving Anthropic's Fable for more difficult expert-level work.

This trend has accelerated since Mozilla's inaugural State of Open Source AI report published in July. The narrowing gap is measured by AI model time horizons, with the best closed models currently able to handle tasks 1.7 times longer than the best open models. If an open model can handle a seven-hour job, a closed model can handle a 12-hour one.

In four months, open models will close this gap entirely. Tasks requiring between eight and 12 hours typically favor closed models, while shorter tasks under eight hours can be effectively handled by cheaper open alternatives. The report notes that closed models retain value due to out-of-the-box functionality, bundled compliance packaging, support, and accountability, which many organizations lack the staff to manage with open models.