Open-Weight AI Models

An open-weight AI model makes its trained numerical parameters, called weights, available for users to download and run. A fully open-source AI model is broader: besides weights, it provides the code and sufficient documentation or training-related materials under licences that permit study, use, modification and redistribution. Because definitions of open-source AI vary, openness should be assessed component by component rather than inferred from a label.

What access to weights enables

Weights encode the parameter values learned during training. Their availability can permit users, subject to licence and technical constraints, to deploy locally, fine-tune the model, conduct independent tests and integrate it without sending every query to the original provider.

• Local operation can provide greater control over data handling, model configuration and continuity of access.
• Users must supply suitable computing infrastructure, security controls, updates and technical expertise.
• Weights alone may not reveal the training data, training process or complete source code.

Open-weight and fully open-source models

Open-weight describes access to one important model component, whereas fully open-source denotes broader technical transparency and legal permission across the model's components.

• An open-weight licence may restrict commercial use, redistribution or particular applications.
• A more fully open model may disclose training and inference code, architecture, documentation and information or materials needed to understand how it was developed.
• The relevant licences should permit users to study, use, modify and redistribute the disclosed components.
• Openness must therefore be assessed across weights, code, data, documentation and licensing; the release of weights alone is insufficient.

How hosted models differ

In a hosted arrangement, the provider operates the model on its own infrastructure and users obtain outputs through an API or interface. The provider controls the underlying weights, deployment environment, safeguards, pricing and updates.

• Users generally cannot inspect, copy or independently modify the parameters of a proprietary hosted model.
• Hosting reduces the user's infrastructure burden but creates dependence on the provider's access rules, service continuity and data-governance arrangements.
• An open-weight model may also be offered as a hosted service, so parameter access and delivery mode are separate questions.