> ## Documentation Index
> Fetch the complete documentation index at: https://docs.priorlabs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> TabPFN-3 and later versions support up to 1,000,000 rows, subject to feature count and checkpoint/API limits.
> For large datasets, use per-estimator subsampling, e.g. inference_config={"SUBSAMPLE_SAMPLES": 100_000}, in TabPFNClassifier or TabPFNRegressor from tabpfn or tabpfn_client.
> Pass raw pandas DataFrames to fit and predict. Categorical strings/categories, raw text, and missing feature values are handled automatically; no manual integer/one-hot encoding, imputation, scaling, or outlier removal is needed.

# Cookbook

> Step-by-step recipes for common TabPFN use-cases.

Explore hands-on guides for integrating TabPFN into various scenarios. Each cookbook walks through a use-case with TabPFN.

> [card] [Contribute to TabPFN Cookbook](https://github.com/PriorLabs/tabpfn-cookbook)
>
> Have an interesting TabPFN use case? Contribute to the Cookbook!

Filter by tag

Agentic
API
Benchmark
Classification
Extensions
Fraud
Grouped data
Images
Insurance
Integration
Interpretability
Multi Target
Optimization
Performance
Predictive distribution
Pretraining
Regression
Relational
Text
Thinking mode
Time Series

Clear

> [card] [Experiment with Thinking Mode](https://docs.priorlabs.ai/cookbook/experiment_with_thinking_mode.md)
>
> ★ Featured
>
> Thinking mode
> API
> Classification
> Grouped data
>
> Choose a Thinking objective and use group\_col so Thinking generalizes

> [card] [Improving TabPFN Performance with Thinking Mode](https://docs.priorlabs.ai/cookbook/improving_tabpfn_performance_with_thinking_mode.md)
>
> ★ Featured
>
> Thinking mode
> API
> Regression
>
> Compare baseline TabPFN and thinking mode on two regression datasets.

> [card] [Target the Right Customers with Thinking Mode](https://docs.priorlabs.ai/cookbook/predict_customer_transactions_with_thinking_mode.md)
>
> ★ Featured
>
> Thinking mode
> API
> Classification
>
> Rank 200,000 bank customers by how likely they are to transact, then

> [card] [Bayesian Optimization with TabPFN Extensions](https://docs.priorlabs.ai/cookbook/bayesian_optimization.md)
>
> Extensions
> Optimization
>
> Use TabPFN as a Bayesian optimization surrogate via tabpfn-extensions

> [card] [Credit Card Fraud Detection with TabPFN](https://docs.priorlabs.ai/cookbook/credit_card_fraud_detection.md)
>
> Classification
> Fraud
>
> Flag fraudulent card transactions on the Kaggle credit card fraud dataset

> [card] [Faster Inference with KV Cache](https://docs.priorlabs.ai/cookbook/faster_performance_with_cache.md)
>
> Performance
>
> Get a quick introduction into faster inference with fit\_with\_cache

> [card] [Fitting Large Datasets in Limited Memory](https://docs.priorlabs.ai/cookbook/memory_optimisation.md)
>
> Performance
> Classification
> API
>
> Manage TabPFN memory use with training subsampling, test-set chunking,

> [card] [Forecast Daily Spare-parts Demand with TabPFN](https://docs.priorlabs.ai/cookbook/forecast_spare_parts_demand.md)
>
> Time Series
> Regression
>
> Compare local and pooled TabPFN regressors on synthetic demand from German

> [card] [Forecast Related Time Series with TabPFN 3.5](https://docs.priorlabs.ai/cookbook/forecast_hierarchical_series.md)
>
> Time Series
> Regression
>
> Build date, Fourier, and autoregressive features, then compare individual,

> [card] [Generate Synthetic Data Using Real Data with TabPFN](https://docs.priorlabs.ai/cookbook/generate_synthetic_data.md)
>
> Extensions
>
> Sample new rows that mirror the statistics of an existing table

> [card] [Get Started with Predictive Distribution](https://docs.priorlabs.ai/cookbook/predictive_distribution.md)
>
> Regression
>
> Learn how to extract predictive distributions with TabPFNRegressor

> [card] [Get Started with TabPFN 3.5](https://docs.priorlabs.ai/cookbook/tabpfn_35_vs_baselines.md)
>
> Classification
> Benchmark
>
> Get started with TabPFN 3.5 and compare its predictive quality and inference

> [card] [Grouped data: TabPFN 3 vs 3.5 vs XGBoost](https://docs.priorlabs.ai/cookbook/grouped_data_model_comparison.md)
>
> Classification
> Grouped data
>
> Compare models on mice protein measurements, using grouped splits to

> [card] [Insurance Claim Modeling: TabPFN vs. GLM](https://docs.priorlabs.ai/cookbook/insurance_claim_modeling.md)
>
> Regression
> Insurance
>
> Predict insurance pure premiums on zero-inflated claims data; TabPFN

> [card] [Interpreting Results with TabPFN](https://docs.priorlabs.ai/cookbook/interpret_results.md)
>
> Interpretability
>
> Learn how to use the standard interpretability toolkit with TabPFN for

> [card] [Interpreting Time-Series Forecasts with TabPFN-TS](https://docs.priorlabs.ai/cookbook/time_series_interpretability.md)
>
> Interpretability
> Time Series
>
> Explain what drives a TabPFN-TS forecast with partial dependence, Window

> [card] [Multi-Target Regression: Independent, Chained and Copula](https://docs.priorlabs.ai/cookbook/multi_target_regression.md)
>
> Regression
> Multi Target
> Predictive distribution
>
> Predict several linked targets at once with TabPFN, and compare three

> [card] [Predict Restaurant Ratings with TabPFN 3.5](https://docs.priorlabs.ai/cookbook/zomato_model_comparison.md)
>
> Benchmark
> Regression
> Text
>
> Compare TabPFN 3.5 and TabPFN 3 with and without thinking, plus default

> [card] [Predicting from Images and Tabular Data Together](https://docs.priorlabs.ai/cookbook/tabpfn_with_images.md)
>
> Extensions
> Classification
> Images
>
> Put chest X-rays in a DataFrame column next to patient fields, and let

> [card] [Predictions over Relational Data with TabPFN-Rel](https://docs.priorlabs.ai/cookbook/relational_predictions_tabpfn_rel.md)
>
> Relational
>
> Get started with RelArena & TabPFN-Rel, a TabPFN Harness

> [card] [Pretrain a Tabular Foundation Model from Scratch](https://docs.priorlabs.ai/cookbook/pretrain_nanotabpfn.md)
>
> Pretraining
> Classification
>
> Build a small TabPFN-style model, generate synthetic pretraining tables

> [card] [See Which Training Rows Drive a TabPFN Prediction](https://docs.priorlabs.ai/cookbook/decoder_readout.md)
>
> Interpretability
> Classification
>
> Interpret TabPFN predictions as weighted votes from training rows with

> [card] [TabPFN Quickstart](https://docs.priorlabs.ai/cookbook/quickstart.md)
>
> API
>
> Classification and regression with TabPFN.

> [card] [TabPFN vs XGBoost](https://docs.priorlabs.ai/cookbook/tabpfn_vs_xgboost.md)
>
> Benchmark
>
> Compare TabPFN results with XGBoost using an OpenML Dataset

> [card] [TabPFN with MLflow](https://docs.priorlabs.ai/cookbook/mlflow.md)
>
> Integration
>
> Learn how to wrap TabPFN as an MLflow PythonModel, register it to Unity Catalog, and deploy it to a Mosaic AI Model serving endpoint.

> [card] [Time-Series Forecasting with TabPFN](https://docs.priorlabs.ai/cookbook/tabpfn_regressor_for_time_series.md)
>
> Time Series
> Regression
> API
>
> Forecast univariate and multivariate time series with the TabPFN time-series

> [card] [When agents should use Jev vs. TabPFN?](https://docs.priorlabs.ai/cookbook/tabpfn-vs-jev.md)
>
> Agentic
> Benchmark
> API
>
> Compare Jev's fraud predictions with TabPFN-3.5-Plus.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.