An open-weight model is an AI model where the trained parameters (the weights) are publicly available to download and run.

In practice, this usually means you can:

  • Run the model locally or on your own cloud.
  • Avoid paying per-request API pricing to a hosted provider.
  • Fine-tune or adapt the model for your own use cases.

However, open-weight does not always mean fully open source. Many open-weight releases do not include all of the following:

  • Full training code.
  • Complete training dataset details.
  • A license that allows unrestricted study, modification, and redistribution.

That is why the distinction between open-weight and open-source matters.

References