Mozilla Report: Gap Between Open and Closed AI Models Narrows to 4.4 Months

Mozilla released its second Open Source AI report, saying the gap between top US closed-source models and China's best open-weight models has narrowed to 4.4 months. The report now uses task duration instead of overall performance scores.

The report highlights Moonshot AI's Kimi K3, which scores just three points below Anthropic's Fable 5 on the Artificial Analysis Intelligence Index, while costing about 30% as much. Open models can handle 7-hour tasks, while closed models handle about 12-hour tasks.

Mozilla CTO Raffi Krikorian says closed models charge a premium mainly for 8-to-12-hour specialized tasks, while open models handle everyday work at very low cost. Companies are already shifting: DoorDash uses Kimi for routine tasks and keeps Fable for complex work.

In August 2026, eight of the top ten models by token usage on OpenRouter were open-source. The report reiterates that open models earn only 4% of revenue, while closed models take 96%.