Skip to main content
By subscribing to TabPFN-2.5 through the AWS SageMaker Marketplace, you can automatically provision and configure TabPFN-2.5 inside your own AWS account - ensuring your data remains within your private AWS networks. TabPFN-2.5 on SageMaker is ideal for teams already operating on AWS who want to benefit from TabPFN-2.5’s performance without managing infrastructure themselves, while maintaining strong data security within their AWS environment:
  • Complete data privacy - Runs in your AWS account; data never leaves your infrastructure.
  • Minimal infrastructure work - AWS handles GPU provisioning and deployment.
  • AWS native - Seamless integration with your existing environment and security policies.
Using TabPFN-2.5 on the AWS SageMaker Marketplace is free of charge; you only pay for the underlying AWS compute. Model weights released under TabPFN-2.5 License. This license is designed to be permissive for research and internal evaluation.For all production use cases, we offer a Commercial Enterprise License. This provides access to our proprietary high-speed inference engine, dedicated support, integration tooling, and other internal models. Please contact us at [email protected] for commercial licensing inquiries.

​
Getting Started

Setting up TabPFN-2.5 in your AWS account is easy and takes just a few steps.
  1. Open the SageMaker Marketplace listing.
  2. Click “View Purchase Options.”
  3. Scroll to the bottom of the page and select “Subscribe”.
After subscribing, Amazon SageMaker may take several minutes to confirm your agreement. This delay is normal, even for free-of-charge products. Next, set up an Endpoint in the AWS Management Console and SageMaker AI.
  1. Navigate to SageMaker AI.
  2. Select the AWS region where you want to deploy TabPFN-2.5.
  3. In the left-hand panel, open “AWS Marketplace resources”, go to the “AWS Marketplace subscriptions” tab, and select TabPFN-2.5.
  4. Click “Actions” on the right-hand side and choose “Create endpoint”.
You will now be prompted to set a Model name and assign an IAM execution role for the model. You can use the creation wizard to streamline this step. After clicking Next, you will be asked to either select an existing endpoint configuration or create a new one. TabPFN-2.5 requires at least one NVIDIA T4 or P4 GPU instance, and for larger datasets we recommend using more capable hardware such as ml.g5.2xlarge or ml.p4.4xlarge for improved performance. Take a moment to confirm that your chosen instance type meets your workload needs before proceeding. The full list of supported machine types for real-time inference and batch transform can be found in the Marketplace listing details page. Once you have clicked on Submit, AWS will automatically set up TabPFN-2.5 in your AWS account and you’re good to go!

Example Code

Step-by-step instructions for running inference with TabPFN-2.5 on SageMaker.

Getting Started Notebook

A guided notebook demonstrating how to use TabPFN-2.5 for inference on SageMaker.

​
Limitations

​
Payload size

SageMaker Models on the AWS Marketplace do not allow any outbound network calls - including calls to AWS-managed services such as S3. As a result, all data must be included directly in the inference request payload, and AWS enforces a 25 MB maximum payload size. TabPFN-2.5 supports two input formats for inference:
  • application/json - a JSON-encoded request body.
  • multipart/form-data - containing the dataset as Parquet files.
Both formats must remain within the 25 MB SageMaker payload limit. Because Parquet is compressed, the multipart/form-data option generally allows you to send more rows or features within the same size constraint.