TabPFN

TabPFN-3 scales to 10M rows. 4M+ downloads, 8k+ GitHub stars, in production from liquid biopsy to rail maintenance. Code: https://github.com/PriorLabs/TabPFN Since July we're an independent lab inside SAP, with more than EUR 1B committed over four years. Models stay open, research stays public, same team and offices. Roles (most can sit in any of the three offices): - Technical Product Manager, Integrations: take every model release live across SAP (AI Core to SAP Analytics Cloud), the cloud marketplaces and customer environments, and help decide which channels we build next. Reports to me, close to the code. - ML Engineer, Infrastructure: own multi-cluster GPU infra (Slurm on GCP today, multi-provider next), training performance and the tooling layer. We spend tens of millions per year on compute; you own that budget. - Research Scientist, Foundation Model: drive the model agenda - novel architectures, scaling from 10K to 1M+ samples, multimodal and causal directions. PhD plus top-venue publications, or equivalent. - Research Engineer, Foundation Model: same agenda from the engineering side. You design experiments, write the training and eval infra, and co-author the papers. - ML Engineer, Backend: design and scale the backend that serves and finetunes the models. Python/FastAPI, Terraform, K8s. - Full Stack Engineer, ML Platform: build the product end to end. TypeScript + Python, React/FastAPI/Postgres. Also hiring: Applied Scientist, Forward Deployed ML Engineer, Research Scientist (Foundational Data Science), PhD research interns, plus GTM and ops roles. 40+ people with backgrounds from Google, DeepMind, Jane Street, Goldman, G-Research, CERN. Led by Frank Hutter, advised by Bernhard Schölkopf and Yann LeCun. Comp competitive with top AI labs. All roles and applications: https://jobs.ashbyhq.com/prior-labs Questions welcome in the replies here.

s to 10M rows. 4M+ downloads, 8k+ GitHub stars, in production from liquid biopsy to rail maintenance. Code: https://github.com/PriorLabs/TabPFN Since July we're an independent lab inside SAP, with more than EUR 1B committed over four years. Models stay open, research stays public, same team and offices. Roles (most can sit in any of the three offices): - Senior ML Infrastructure Engineer: own multi-cluster GPU infra (Slurm on GCP today, multi-provider next), training performance and the tooling layer. We spend tens of millions per year on compute; you own that budget. - Research Scientist, Foundation Model: drive the model agenda - novel architectures, scaling from 10K to 1M+ samples, multimodal and causal directions. PhD plus top-venue publications, or equivalent. - Research Engineer, Foundation Model: same agenda from the engineering side. You design experiments, write the training and eval infra, and co-author the papers. - ML Engineer, Cloud Platform: design and scale the backend that serves and finetunes the models. Python/FastAPI, Terraform, K8s. - Full Stack Engineer, ML Platform: build the product end to end. TypeScript + Python, React/FastAPI/Postgres. Also hiring: Applied Scientist, Forward Deployed ML Engineer, Research Scientist (Foundational Data Science), PhD research interns, plus GTM and ops roles. ~40 people with backgrounds from Google, DeepMind, Jane Street, Goldman, G-Research, CERN. Led by Frank Hutter, advised by Bernhard Schölkopf and Yann LeCun. Comp competitive with top AI labs. All roles and applications: https://jobs.ashbyhq.com/prior-labs Questions welcome in the replies here.

000 × 500 in v2) - SOTA performance: TabPFN-2.5 outperforms tuned tree-based methods and matches the performance of a complex ensemble (AutoGluon 1.4), that itself includes TabPFN v2, tuned for 4 hours. Tuning the model improves performance, outperforming AutoGluon 1.4 for regression tasks. - Rebuilt API: New REST interface along with Python SDK with dedicated fit & predict endpoints, making deployment and integration more developer-friendly - A distillation engine that converts TabPFN-2.5 into a compact MLP or tree ensemble while preserving accuracy and offer low latency inference. There are still some limitations. The model is designed for datasets up to 50K samples. It can handle larger datasets but that hasn’t been our focus with TabPFN-2.5. The distillation engine is not yet available through the API but only through licenses (though we do show the performance in the model report). We’re actively working on removing these limitations and intend to release newer models focused on context reasoning, causal inference, graph networks, larger data and time-series. TabPFN-2.5 is available via API and a package on Hugging Face. Would love for you to try it and give us your feedback! Model report: https://priorlabs.ai/technical-reports/tabpfn-2-5-model-repo... Package: https://github.com/PriorLabs/TabPFN Client: https://github.com/PriorLabs/tabpfn-client Docs: https://docs.priorlabs.ai/quickstart

https://discord.com/invite/VJRuU3bSxt Would love to hear what you think when you try it!

set acts as one meta-datapoint to train the TabPFN weights with SGD. As a foundation model, TabPFN allows for fine-tuning, density estimation and data generation. Compared to TabPFN v1, v2 now natively supports categorical features and missing values. TabPFN v2 performs just as well on datasets with or without these. It also handles outliers and uninformative features naturally, problems that often throw off standard neural nets. TabPFN v2 performs as well with half the data as the next best baseline (CatBoost) with all the data. We also compared TabPFN to the SOTA AutoML system AutoGluon 1.0. Standard TabPFN already outperforms AutoGluon on classification and ties on regression, but ensembling multiple TabPFNs in TabPFN v2 (PHE) is even better. There are some limitations: TabPFN v2 is very fast to train and does not require hyperparameter tuning, but inference is slow. The model is also only designed for datasets up to 10k data points and 500 features. While it may perform well on larger datasets, it hasn't been our focus. We're actively working on removing these limitations and intend to release new versions of TabPFN that can handle larger datasets, have faster inference and perform in additional predictive settings such as time-series and recommender systems. We would love for you to try out TabPFN v2 and give us your feedback!