Editorial illustration for Mozilla Report: Open AI Models Should Be Default for Most Work
Open AI Models Now Match Proprietary Systems: Mozilla
Mozilla Report: Open AI Models Should Be Default for Most Work
Four months and change. That's how long US tech companies' priciest AI models can now hold a performance lead over open-weights models built in China, according to a new report from Mozilla. The gap used to be a moat. Now it's closer to a speed bump, and it's forcing a lot of companies to ask why they're still paying five times as much for a head start that keeps shrinking.
Mozilla's State of Open Source AI report, published September 15 and shared with Ars ahead of release, argues that most organizations should treat open models as their default choice for the bulk of everyday AI work. Moonshot AI's Kimi K3, an open model, scores just three points behind Anthropic's closed Fable 5 on the Artificial Analysis Intelligence Index, a widely used composite benchmark, while costing roughly 30 percent as much to run.
That price gap is why the calculation companies are making has shifted from "which model is best" to "which workloads actually justify the premium." Mozilla's CTO, Raffi Krikorian, laid out where that premium still holds up, and where it doesn't.
The performance gap between frontier AI models from US tech companies and the best open-weights models from Chinese companies has closed to just 4.4 months, according to a Mozilla report. That explains why many companies are shifting to the significantly cheaper open models for routine work—and helps reveal a narrow band of workloads where frontier models are worth the cost.
Why this matters
For anyone building products, that 4.4-month gap is a procurement number now, not a research trivia point. If Moonshot's Kimi or similar open-weights models cover 90% of routine tasks at a fifth of the cost, paying frontier prices for everything is just bad budgeting. We'd push teams to actually map their workloads: which tasks need the newest OpenAI or Anthropic model's edge, and which are fine on an open model running on your own infrastructure.
Mozilla's framing, open models as default and frontier as the exception, flips the usual enterprise instinct to buy the best available. That's worth taking seriously, especially with Chinese labs setting the open-weights pace. But default doesn't mean automatic.
Licensing terms, data residency, and support commitments vary a lot between open projects, and "closed the gap" isn't the same as "closed for every task." The real work now is figuring out where that narrow band of frontier-worthy jobs actually sits inside your own stack, rather than assuming either extreme. Watch how fast that 4.4-month window moves in either direction over the next few quarters.
Related Reading
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
From Archives Agent-Net Open Sources Go Harness for Web-Based AI Agents
From Archives AI Bots Timmy, Ren, and Jackie Flood Social Media With Slop
Common Questions Answered
According to Mozilla's report, what is the current performance gap between US frontier AI models and Chinese open-weights models?
According to Mozilla's State of Open Source AI report published in September, the performance gap has narrowed to just 4.4 months, meaning Chinese open-weights models can match the capabilities of US tech companies' priciest AI models within that timeframe. This represents a significant reduction from the previous wider gap that existed between these two categories of models.
Why are companies shifting away from expensive frontier AI models to open-weights models for routine work?
Companies are making this shift because open-weights models now cover approximately 90% of routine tasks while costing significantly less—about a fifth of the price of frontier models from companies like OpenAI and Anthropic. With the performance gap closing to just 4.4 months, paying premium prices for frontier models on all tasks represents poor budgeting when cheaper alternatives can handle most standard workloads effectively.
What does Mozilla recommend companies do to optimize their AI model procurement strategy?
Mozilla recommends that teams map their specific workloads to identify which tasks truly require the newest edge from frontier models like OpenAI or Anthropic, and which tasks can be adequately handled by open-weights models running on their own infrastructure. This strategic approach allows companies to use frontier models only where their performance advantage is necessary while leveraging cheaper open models for routine work to improve overall budgeting efficiency.
How has the competitive landscape between US and Chinese AI models changed according to this Mozilla report?
The report indicates that what was once a significant competitive moat for US tech companies has now become merely a speed bump, with Chinese open-weights models closing the performance gap to just 4.4 months. This narrowing gap is forcing companies to reconsider their reliance on expensive US frontier models and is fundamentally changing procurement decisions across the industry.
Further Reading
- Mozilla’s Inaugural ‘State of Open Source AI’ Report Is Here | The Mozilla Blog - Mozilla Blog
- China captured the global lead in open-weight AI development during 2025, Stanford analysis shows - The Decoder
- China's open AI models are in a dead heat with the West - here's what happens next - ZDNET
- Mozilla Challenges Monopolies With Inaugural State of Open Source AI Report - Open Source For You
- Mozilla report claims open-source AI nearly matches Big Tech models - Northeast Times
