Skip to main content
TabPFN-2.5 and TabPFN-2.6 are released under a non-commercial license, which you need to accept before the model files can be downloaded. On first use, TabPFN will automatically open a browser window where you can log in via PriorLabs and accept the license terms. Your authentication token is cached locally, so you only need to do this once. 
from tabpfn import TabPFNClassifier

# This will open a browser for login and
# license acceptance on first use
model = TabPFNClassifier()
model.fit(X_train, y_train)

​
Notebooks

If you run TabPFN via a Notebook, the browser-based setup will not be available. Instead, you can follow these steps:
  • Visit https://ux.priorlabs.ai and log in
  • Go to the Licenses tab and accept the license
  • Go to the API Keys tab and copy your API key
  • In your notebook:
import os
os.environ["TABPFN_TOKEN"] = "<your API key here>"

from tabpfn import TabPFNClassifier

# On first use, this will automatically download
# the model weights using the TABPFN_TOKEN set above
model = TabPFNClassifier()
model.fit(X_train, y_train)

​
Google Colab

In Colab, it’s recommended that you set the TABPFN_TOKEN secret so that you don’t have to go through the license flow each time:
  • Visit https://ux.priorlabs.ai and log in
  • Go to the Licenses tab and accept the license
  • Go to the API Keys tab and copy your API key
  • In Colab, select “Secrets” from the left-hand menu
  • Create a new secret with name TABPFN_TOKEN and value set to your API key
  • Ensure that “Notebook access” is enabled

​
Manual Setup

For environments where neither a browser nor outbound internet connectivity for downloading the model weights is available:
  • Visit https://ux.priorlabs.ai and log in
  • Go to the Licenses tab and accept the license
  • Click “Model Weights” to download the weights. Place them into a folder /path/to/tabpfn-weights
  • Set the environment variable export TABPFN_MODEL_CACHE_DIR=/path/to/tabpfn-weights
  • Run your script:
from tabpfn import TabPFNClassifier

# This will respect the TABPFN_MODEL_CACHE_DIR environment variable
# and not make any outbound network calls for downloading the weights.
model = TabPFNClassifier()
model.fit(X_train, y_train)
You can also set TABPFN_NO_BROWSER to disable the automatic browser login if needed (e.g. in environments where opening a browser is undesirable). If access via any of these options is not an option for you, please contact us at sales@priorlabs.ai.