Token cost
Each prediction request is charged based on the number of rows, columns, and estimators used. TabPFN v3 uses a sublinear cost function — doubling your data does not double the cost. Because TabPFN-3 is fundamentally more efficient at handling larger datasets, large-scale workloads are significantly cheaper than on v2.x. TabPFN v2.x uses a linear cost function:n_estimators defaults to 8 when not specified. The floor of 5,000 tokens applies to every request regardless of size.
You can check your current usage with the get_api_usage() function in tabpfn-client, or monitor it in your account dashboard at ux.priorlabs.ai.
Usage pools
When a pool is exhausted, the API returns HTTP 429 with a message indicating which limit was hit and when it resets.
A warning email is sent when either daily or monthly usage crosses 70% of the limit.
Thinking fit quota
Thinking mode fits are metered separately from prediction tokens. Each call toPOST /tabpfn/fit with thinking enabled counts against the monthly thinking fit quota, regardless of dataset size or fit duration.
When the thinking fit limit is reached, /tabpfn/fit returns HTTP 429. Prediction requests are unaffected — you can still predict using previously fitted models.
Dataset limits
Limits vary by model version. UseGET /tabpfn/get_model_limits to retrieve the current values for your account.
Regression full output:
output_type="full" is limited to 400 test rows. Use "mean", "median", or "quantiles" for larger test sets.
HTTP error codes
Higher limits
If you need higher token budgets or more thinking fits, contact hello@priorlabs.ai with your projected daily token volume and peak thinking usage.Thinking mode
Configure fit-time optimization and understand thinking parameters.
REST quickstart
Full upload → fit → predict walkthrough.
TabPFN-3 changelog
What’s new in v3 — scale, capabilities, and migration.
Security
Encryption, data isolation, and access controls.