Skip to main content
Whether you’re just getting started with TabPFN or pushing it into production, this FAQ highlights practical answers to common questions about model limits, performance, reproducibility, and API best practices.
At minimum, TabPFN requires an NVIDIA T4 GPU to run efficiently. For best performance, we recommend A100 or H100 GPUs. CPU and MPS (Apple Silicon) execution are supported experimentally, but GPUs provide the best throughput for production workloads and larger tables.
TabPFN is optimized for small- to medium-sized tabular contexts, since it performs in-context learning using transformer attention. See our Models for an overview of capabilities.
In the local package version text features are encoded as categoricals without considering their semantic meaning. Our API automatically detects text features and includes their semantic meaning into our prediction. The local package version encodes text as numerical categories and does not include semantic meaning.
In the local package version date features are encoded as categoricals without considering their semantic meaning. Our API automatically detects date features and creates an optimized embedding.
With a fixed seed and in the same environment TabPFN inference is deterministic. Across different hardware (CPU, GPU, MPS) configurations small differences are expected.
Yes. TabPFN’s estimators can handle missing values internally, including pd.NA, without requiring manual imputation.
By default TabPFN performs most of its fitting inside the predict() step (that’s how it simulates Bayesian posterior inference). That means latency scales roughly with the number of training rows. For fast inference we can create a tree- or small MLP based model that yields almost the same accuracy as TabPFN. Contact sales@priorlabs.ai to access this solution.
The current API rate limits are 100 million cells per day, where a “cell” is computed as:
api_cost = max((num_train_rows + num_test_rows) * num_cols * n_estimators, 5000)
These limits automatically reset every 24 hours. If you require further credits, please fill out this form.
After making predictions with a TabPFN model using the client you can access model-level metadata, such as the exact model version used for inference. Each estimator exposes a .last_meta attribute containing this information:
# After model.predict():
meta = model.last_meta
print(meta)
# Example output:
# {‘model_version’: ‘v2’, ‘info’: ’Mi4yLjB8MjAyNS0xMC0zMVQxMjowMTo0My42Mzc3ODcrMDA6MDA=’}