Open-weight model

An AI model whose trained parameters are published, so anyone can download, run and modify it.

• The weights are the billions of numbers learned in training; with them and standard software, the model runs on your own hardware.
• It is not full open source: training data and code are usually withheld, so others can use and adapt the model but not rebuild it.
• Benefits are cost control, data privacy and research access, and cheap fine-tuning for narrow tasks.
• Release is irreversible, and safety training can be removed by fine-tuning, which is the core of the policy debate.