Training Custom Models via the API

You can easily fine-tune your own custom model using our API. The training process is divided into 2 steps:

  1. Upload a dataset crafted specifically to your use-case.
  2. Train your own custom model using that dataset.

    This process can be done in the AI21 Studio website or by posting an HTTP request to the relevant endpoint.

Fine-tuning is currently supported for J2-Grande and J2-Large, as well as our legacy models J1-Jumbo, J1-Grande, and J1-Large.