There's a quiet, good thing happening in the AI world right now, and we think it's worth naming out loud. Open weight models — the kind of AI you can download, study, and run yourself — are bringing a real balance to the ecosystem. They're not replacing the closed labs. They're not winning a race. They're doing something more important: making intelligence accessible to anyone with a laptop and a weekend. We think that's a genuinely good thing — not just for AI, but for the world.
What's an open weight model, in plain language
An AI model is, at heart, a very large set of numbers — the "weights" — that determine how it thinks. Open weight means those numbers are public. Anyone can download them, run them on their own hardware, study them, fine-tune them, and build on top without asking permission or paying per-message fees. Closed weight means the numbers stay on the lab's servers, and the rest of us only ever get the answers back.
Both kinds exist for a reason. The closed labs have spent billions training frontier models, and the closed approach is theirs to choose. The open labs are doing different work — releasing the weights, building in the open, and betting that the capability will be more useful if it isn't locked up. Neither side is going away. Both are good.
The balance they're bringing
For a while, the world had a single shape: a few big AI labs, mostly closed, mostly expensive, mostly out of reach for anyone outside a well-funded company. That's a healthy thing to question, and the open weight movement has been the answer to that question.
The balance it brings isn't a competition where one side wins. It's the simple presence of an alternative. A small bakery in Auckland and a large retailer in Berlin can use the same underlying intelligence. A research lab in Nairobi and a startup in São Paulo can fine-tune it for their language, their domain, their community. The capability doesn't sit behind someone else's pricing page or someone else's terms of service. It belongs to anyone willing to do the work to use it.
The open weight labs are doing that work. We benefit from it, and so does the rest of the world.
Where Ullbek stands
We are a customer of all these labs. We pick the best model for the work, regardless of whether the weights are open or closed. We support the open labs because their work makes our product better, and we support the closed labs because their work is often the right tool for the job.
The default model on Ullbek happens to be open weight right now, and we're grateful to the people who built it. If a closed model takes the seat back next year, we'll make that change too — the same way we've made every other change when the math moved. The point isn't to pick a side. The point is to keep using the best of what's available, in either world, for the people building with us.
That's our take
The AI world is healthier with both ecosystems in it. The competition between them is good for everyone. The existence of the open alternative is good for the world in a way that goes beyond product roadmaps. We're glad the open weight movement is doing this work. We'll keep using what they build. We'll keep telling you which one we picked and why.