Fast model (α)

TabPFN-3.5-Fast (alpha) is a pretrained checkpoint designed for faster inference. It uses the familiar fit and predict workflow for classification and regression. Fast is available through the local tabpfn package, the hosted API and tabpfn-client, and VPC deployments. See Models for limits and availability across the TabPFN-3.5 family.

Use the fast checkpoint

These examples assume you have prepared X_train, y_train, and X_test for your task.

API (tabpfn-client)
OSS (tabpfn)

Use tabpfn-client to run TabPFN-3.5-Fast on the hosted API. Follow the quickstart for installation and API key setup.Select model_path="v3.5-fast_default" when creating the estimator.

from tabpfn_client import TabPFNClassifier

model = TabPFNClassifier(model_path="v3.5-fast_default")
model.fit(X_train, y_train)
predictions = model.predict(X_test)

from tabpfn_client import TabPFNRegressor

model = TabPFNRegressor(model_path="v3.5-fast_default")
model.fit(X_train, y_train)
predictions = model.predict(X_test)

Follow the quickstart to install a tabpfn release with TabPFN-3.5-Fast support. The package downloads and caches the weights automatically; see Accessing model weights for license acceptance and authentication.Select ModelVersion.V3_5_FAST when creating the estimator.

from tabpfn import TabPFNClassifier
from tabpfn.constants import ModelVersion

model = TabPFNClassifier.create_default_for_version(ModelVersion.V3_5_FAST)
model.fit(X_train, y_train)
predictions = model.predict(X_test)

from tabpfn import TabPFNRegressor
from tabpfn.constants import ModelVersion

model = TabPFNRegressor.create_default_for_version(ModelVersion.V3_5_FAST)
model.fit(X_train, y_train)
predictions = model.predict(X_test)

See Selecting Model Version for more on choosing models in the local package and hosted API.

Getting Started

Metering and rate limits

Capabilities

Agentic

Integrations

Troubleshooting

TabPFN in Action

Customer References

Changelog

Use the fast checkpoint

​Use the fast checkpoint

Use the fast checkpoint