AI News Feed
Market watch
TechRadar Showcases Wildlife Photos Shot With Sony 400mm and 600mm Super-Telephoto PrimesREI Waterproof Jacket Deals Offer Up to 50% Off Patagonia, Arc'teryx and MoreGoogle ATLAS Reports Global AI Use Patterns and Scientist Time SavingsAI Pioneers Add to Warnings of Catastrophic Risk as Washington Splits on RegulationSimpliSafe's Video Doorbell Series 2 adds human Active Guard monitoringCurved monitors deliver immersive gaming but carry trade-offs, Engadget says
TechRadar Showcases Wildlife Photos Shot With Sony 400mm and 600mm Super-Telephoto PrimesREI Waterproof Jacket Deals Offer Up to 50% Off Patagonia, Arc'teryx and MoreGoogle ATLAS Reports Global AI Use Patterns and Scientist Time SavingsAI Pioneers Add to Warnings of Catastrophic Risk as Washington Splits on RegulationSimpliSafe's Video Doorbell Series 2 adds human Active Guard monitoringCurved monitors deliver immersive gaming but carry trade-offs, Engadget says
Large Language Models

Mozilla Report Puts U.S. Frontier AI Lead at 4.4 Months Over Top Chinese Open Models

Mozilla's State of Open Source AI report, published September 15, says U.S. frontier models lead the best Chinese open-weights models by 4.4 months, and that most organizations should make open models their default.

AI News Feed · 2026-09-15 23:19

The assessment appears in Mozilla's State of Open Source AI report, which was shared with Ars Technica before publication. Mozilla's stated guidance is that most organizations should ideally be using open models for the majority of their work, and that paying for closed frontier models is justified only for a narrow band of workloads.

The report points to Moonshot AI's Kimi K3, which it identifies as a leading open model. On the Artificial Analysis Intelligence Index, Kimi K3's composite performance score sits three points below Anthropic's Fable 5, a closed frontier model, while costing 30 percent of what the closed model costs.

Raffi Krikorian, chief technology officer at Mozilla, described the trade-off as tied to specific tasks. "Closed earns its premium in a few places: expert professional work, high-intensity retrieval, and long context," he said in an email to Ars Technica. "We see the decision to pay for closed as workload-specific rather than organization-specific."

The cost numbers reported by Mozilla differ from the framing in Ars Technica's own headline on the report, which describes paying for frontier models as buying a four-month head start at five times the cost. The Kimi K3 comparison in the report puts the open model at 30 percent of the closed model's price.

According to the report, the 4.4-month gap helps explain why many companies are moving to significantly cheaper open models for routine work, and it also identifies the limited set of workloads in which frontier models are worth their price.

The report's performance comparison draws on the Artificial Analysis Intelligence Index, which combines results across multiple evaluations into a single composite score. Mozilla published the report on September 15 and provided it to Ars Technica in advance.

【Editor's Summary】Mozilla's September 15 report puts the performance gap between U.S. frontier AI models and the best Chinese open-weights models at 4.4 months, with Moonshot AI's Kimi K3 scoring three points behind Anthropic's Fable 5 at 30 percent of its cost. Mozilla advises most organizations to default to open models, reserving closed frontier models for expert professional work, high-intensity retrieval and long-context tasks.