Run TabPFN where
your data already is.

Pre-trained tabular foundation models for making predictions on structured data.
Get best-in-class results in seconds - no tuning, no ML pipelines
Start free

From APl to on-prem, we’ve got it covered

Open weights

For experiments, evaluation and research on your own hardware. Open-source and built with community.
INCLUDED
  • TabPFN-3.5 and Fast
  • tabpfn Python package
  • Non-commercial, open-source license
  • CPU and MPS support for small datasets
  • GPU recommended
Get Started

API

For production, experiments and research. Our most powerful models on EU servers.
INCLUDED
  • TabPFN-3.5-Plus, Fast and Thinking
  • Python SDK, REST API and MCP
  • Use allowed under Terms of Service
  • Only Python and internet connection needed
  • Data processing in EU
Get an API key

Private Cloud

For production, experiments and research. Our most powerful models on your cloud infrastructure.
INCLUDED
  • TabPFN-3.5-Plus, Fast and Thinking
  • Your data never leaves your infrastructure
  • Azure AI Foundry, AWS SageMaker and more
  • Same Python SDK as for our API
Choose platform

On-prem container

For production power users. Our models on your airgapped environment.
INCLUDED
  • TabPFN-3.5 model family
  • Your data never leaves your infrastructure
  • Same Python SDK as for our API
Request container

Same interface, everywhere

YOUR CODE
tabpfn-client

One import, one call

from tabpfn_client import
TabPFNClassifier

model = TabPFNClassifier(
    endpoint="your-vpc-endpoint",
)
model.fit(X_train, y_train)
predictions = model.predict(X_test)
API
Data processing in EU
VPC
AWS, Azure or your pick
On-prem
Your hardware
GA
AWS
SageMaker Marketplace
Get Started
GA
Azure
AI Foundry
Get Started
GA
MCP
EU inference
Get Started
GA
SAP
AI Core
Get Started
EARLY ACCESS
Databricks
Agents & MLFlow
get Started
ON REQUEST
Snowflake
Marketplace
Request access
ON REQUEST
GCP
Vertex AI
Request access

Need TabPFN somewhere else?

Choose what fits you best

Managed API
Private Cloud
On-prem container
INFRASTRUCTURE & SECURITY
Infrastructure needed
A laptop
GPUs, min. T4, H100 recommended
GPUs, min. T4, H100 recommended
Where inference runs
Prior Labs servers; EU
Your hyperscaler
Your airgapped environment
Internet required
Certifications
SOC2 (observation), GDPR
Your certifications
Your certifications
SLA
MODELS Available
Model updates
On request
Base
Plus
Thinking
On request
Fast
On request
Get An API Key
Choose platform
Request

Deploy anywhere

Questions

I don't see my deployment environment in the list. How do I know if you support it?

If you don't see your deployment environment in the list, reach out to us at sales@priorlabs.ai or fill out a form on this page and mention the environment there.

I would like to test Thinking, but cannot send data to the API. What should I do?

Thinking mode is available on Private Cloud, including testing. Contact us at sales@priorlabs.ai to get access.

What is the pricing for each of these deployment options?

You can check out our pricing page or reach out to us at sales@priorlabs.ai

Are the published weights and API / Private Cloud the same, including licence?

We publish our TabPFN model line under an open-source non-commercial licence. API, as well as Private Cloud, is running the Plus and Thinking model lines, which are covered by our Terms of Service.

What is the difference between Plus, Thinking, Fast and Base?

Thinking is our most powerful offering. Plus improves on the base model by adding intelligent text handling, and Fast trades off accuracy for x6 faster inference than base. You can read more about our different model lines on docs.priorlabs.ai.

Tell us where you're building

TabPFN comes to you, not the other way around.

close icon
Request Access Form

Request Access

* indicates required