About this guide

Contents

1. Who it is for
2. What it covers
3. How it is made
4. Corrections

Most pages that rank for "best open source AI" are leaderboards. They say which model scored highest last month, but rarely what its licence allows, what it takes to run or how a company would make it its own. This guide tries to answer those questions.

Who it is for

Chief technology officers, heads of engineering and data, and the executives who ask them whether the company should do what Revolut did. We assume readers understand their own systems and want facts they can check rather than a list of model names.

What it covers

The main guide compares open-weight model families and the layer that turns them into a company’s own model. The supporting guides cover what self-hosting took at Revolut, how to train a model on company data, and what "open source" actually means in a model licence. A European section covers the models built in Europe, the AI Act and where data is held, and is published in German, French, Spanish, Italian, Dutch and Polish as well as English.

How it is made

Every profile is built from model cards, licence files, vendor announcements and reputable reporting. Each fact carries a numbered source and the date we opened it. The comparison uses the same published criteria for every option; they are set out on the methodology page.

We update the main guide at least once a quarter and whenever a family releases a new flagship or changes its licence. Every page shows when it was last updated and keeps a short log of changes.

Corrections

If you think a ranking is unfair or a fact is wrong, tell us. We correct errors quickly and note the change on the page.

Further reading

• Case analysisGoing headless like Revolut: what replacing Claude Code with an open model tookRevolut now runs most of its AI coding on an open-weight model it serves itself, for about a tenth of its former bill. A reconstruction of what it built, what the numbers mean for other companies, and the step that comes after.
• Practical guideHow to train your own AI model on company dataMost of the work in training a company model happens before any training starts. A step-by-step account of the data work, the choice of base model, the training methods and the rules that apply in Europe.
• Licence guideOpen source vs open weights: what each licence lets a company doAlmost every model sold as open-source AI publishes its weights but not its training data. The difference rarely matters for internal use. It matters a great deal for companies that resell a model or build a product on it.

Also in: Deutsch · Français · Español · Italiano · Nederlands · Polski