About ai-opensource.com

ai-opensource.com is a multilingual guide to open AI models: plain-language explanations for beginners, and source-checked specs, licenses and hardware numbers for professionals.

  • Updated

In short. ai-opensource.com keeps a source-checked catalog of open-weights AI models and explains the ideas around them in plain language, in more than one language.

Why this site exists

New open AI models appear almost every week. The announcements are noisy, the licenses differ in ways that matter for business use, and the hardware you need ranges from a phone to a rack of GPUs. We collect the models in one place, state each fact once, and link it to the primary source it came from.

Who it is for

  • Beginners. Every model page starts with a short explanation in plain words, the glossary defines the jargon, and the guides answer the first questions people ask, such as what an open-source AI model is and how much memory you need.
  • Practitioners. Model pages list parameters, active parameters, context length, shipped file sizes and precision formats, plus license conditions and links to the model card and license file. The numbers are also available as open data.

What you will find

  • Models: the catalog, with search, filters for category, license and the hardware you have, and sorting.
  • Best-of lists and comparisons: editorial picks for common needs, with the reasons written out.
  • VRAM calculator: estimate the memory a model needs at different precisions.
  • Licenses: each license in plain language, with what you can do and what to watch for.
  • Guides and the glossary: background knowledge, from first principles to fine-tuning.
  • Tools and developers: the software that runs the models and the teams that build them.
  • News: releases and short analyses, each with sources.

What “open” means here

Most models on this site are open-weights: the trained model files can be downloaded, run on your own hardware and, usually, adapted. That is not always the same as open source in the strict sense. We show the license of every model and mark the few that also publish their training data and code as fully open. Open weights vs open source explains the difference.

How we work

The methodology page explains where facts come from, how we estimate memory, how licenses are classified and how often pages are updated. Every page shows when it was last updated. Text and data are published under CC BY 4.0, so you may reuse them with credit.

Editorial independence

Developers cannot pay to be listed, ranked or described favorably. Benchmark numbers reported by a developer are always labeled as the developer’s own.

Languages

English is the reference version, and translations follow it. A page that has not been translated yet appears in lists with an EN label and opens in English.

Corrections

If you find an error, please tell us through the contact link in the footer. We correct the page and update its date.