Open-Weight Models

Working knowledgeCore AI Concepts

Also called: Open Weights, Open-Weight AI

Open-weight models are models whose parameters are published for anyone to download, inspect, fine-tune or run on their own infrastructure. The distinction from open source is that most open-weight releases do not publish training data or code, only the trained model itself. 2026 saw open-weight models close in on frontier performance at a fraction of the cost.

In practice

An open-weight model run on infrastructure you control can satisfy data residency requirements a closed API never could, which is often the deciding factor in regulated sectors. The trade is that you take on the hosting and the security of the deployment, so price the infrastructure alongside the licence saving.

Not sure where your organisation stands?

Take the free AI-readiness diagnostic.

Start the diagnostic