The state of
open source AI.

Build with us.

Right now, in New Zealand's far north, a Māori broadcaster trains speech models for te reo — a language too small for any market — on open tools, under a license that keeps the data with its people. At PwC, one of the world's biggest accounting firms, teams fine-tuned a small open model on the language of finance and run it today for hundreds of clients — on their own hardware, with no per-token meter running. In East Africa, farmers diagnose cassava disease with a free app whose model runs on the phone itself, offline, in fields the cloud has never reached. In Lausanne, researchers built an open medical model with the Red Cross, tuned to its humanitarian guidelines, with clinical trials being prepared at home and in Tanzania. In Switzerland, a public consortium trained a national model on public supercomputers and released all of it — weights, data, training code — free for any country to copy. They built it themselves. They didn't ask permission, they didn't rent it — they own it, theirs to run, change, and keep

They are not outliers. Open-source and open-weight AI have become one of the fastest-growing builder ecosystems in the history of software: on Hugging Face alone, 2.5 million public models, thirteen million users, a third of the Fortune 500 among them. And the market has voted. On OpenRouter, where developers route real production traffic, open-weight models grew from a sliver to roughly a third of usage by late 2025. Six months on, the platform moves 25 trillion tokens a week — five times more — and the single biggest source of that traffic is an open model. Nobody does that out of idealism. The models are good and the economics are better.

Technology's best-kept secret is publicly available for all to use. This spring, the best closed model scored 60 and the best open models 54. A year earlier, the leading open model scored 22. The frontier still leads on the hardest problems — and for the work most builders actually ship, where price, control, and deployability decide, the data does not say “promising.” It says “ready.” If you have been waiting for open AI to grow up, stop waiting.

Governments are moving too: the European Commission has proposed an “open source first” rule for how public institutions buy AI, and Canada has set a national target to take business adoption from 12 percent to 60. When communities, markets, and states reach for the same thing at once, that is not a movement but a direction — and the direction is abundance: more intelligence, in more hands, owned by more people, than any closed system could distribute.

But none of this is inevitable, and the other future on offer is seductive: a handful of validation machines reading the world to you — smooth, confident, sourced to nothing you can check — the bazaar of a billion arguing voices traded for one polished concierge that answers only to its owner. We got the preview this June, on a Friday afternoon, when one of the most advanced AI models in commercial use went dark everywhere because one government sent one letter. Every business renting it learned what renting means: the off switch was never theirs. But the exit showed itself in the same week, as open-weight models became the instant second source — because no order, from any capital, can un-download weights already running on your own hardware.

We have been here before. Mozilla exists because one company once tried to own the web's front door, and an open community made sure it never would. This is not a new fight; it is the sequel. Then the prize was the front door. Now it is the engine. We bet on open the first time. Open won. Together, we can do it again.

Our belief is simple: the path forward is competition and interoperability — many models, standard connections, and the freedom to leave. Open did not make the pie smaller and share it. It made the pie bigger and let more people own a slice.

So read what follows as a map, not a brochure: where open AI is winning — some numbers surprised even us — and where it is exposed. A case that hides its weak points is an advertisement.

The builders are already building. The motors are turning. A rented future has deeper pockets; an owned one has more hands — millions more — and this story ends the same way every time it is told: the many, building in the open, outbuild the few behind walls.

Build with us.