Frank Hutter

The tabular data science revolution is really here now! I’m thrilled to introduce *TabPFN Scaling Mode*, removing TabPFN’s previous limit on dataset size. The plot below shows scaling results for the biggest dataset we have tried so far (subsamples of 100k and 1M, and the full 10M rows / data points). While we’ve only benchmarked up to 10M rows (a 100x increase in the 3 weeks since the release of TabPFN-2.5!), the results indicate that we can just scale to any dataset size now 🚀 🧵1/4

This is a special moment for us at @prior_labs. I’ve followed @ylecun’s work since the start of my career. His relentless focus on "what’s next" in AI has always been an inspiration. Today, I’m proud to say that he is joining us as a scientific advisor to help build the future of Tabular Foundation Models. Yann understands that while LLMs have captured the world's attention, they have severe limitations. The vast majority of the world's data is tabular and LLMs are terrible with statistics and numbers. Our model, #TabPFN, fills this void, and the progress we are seeing is exponential. A warm welcome to the team, Yann. It’s an honor to work with you alongside @bschoelkopf, Madelon Hulsebos, and @SamuelMullr on this powerhouse board. #DeepLearning #DataScience #TabularData #AI

4/4 We’re currently working directly with customers to run TabPFN-2.5 Scaling Mode on their data. If you’re interested, please contact sales@priorlabs.ai to get on the waitlist.

3/4 More details in our blog post: priorlabs.ai/technical-repo…

The tabular data science revolution is really here now! I’m thrilled to introduce *TabPFN Scaling Mode*, removing TabPFN’s previous limit on dataset size. The plot below shows scaling results for the biggest dataset we have tried so far (subsamples of 100k and 1M, and the full 10M rows / data points). While we’ve only benchmarked up to 10M rows (a 100x increase in the 3 weeks since the release of TabPFN-2.5!), the results indicate that we can just scale to any dataset size now 🚀 🧵1/4

Update: TabPFN-2.5 is actually the SOTA model on all of TabArena (which has datasets with up to 100k training data points). In a single forward pass, TabPFN-2.5 outperforms all other models, even if you tune them for 4 hours. We built and previously evaluated TabPFN-2.5 for up to 50k data points (and 2k features) and were kind of surprised that it's SOTA up to 100k 🙂 👉 TabPFN-2.5 webinar tomorrow: app.livestorm.co/p/21526e44-406… 👉 Model report on arXiv: arxiv.org/pdf/2511.08667

6/6 Please see the TabPFN-2.5 model report for details: priorlabs.ai/technical-repo… API: docs.priorlabs.ai/quickstart #tabpfn #priorlabs

5/ I'm really surprised by the performance. We only aimed for a fairly incremental improvement with this release, but our fantastic team still delivered a step change. Big shoutout to all my amazing colleagues at @prior_labs!

The data science revolution continues — TabPFN is now SOTA up to 50k data points and 2000 features 🚀 For the size limits of TabPFNv2, in a forward pass Real-TabPFN-2.5 outperforms AutoGluon 1.4 (complex ensemble including TabPFNv2 tuned for 4h) by 93 ELO points on TabArena.🧵1/

A new version of TabPFN will come out soon ... Our Discord community will get early access first. Ready to join the beta? discord.com/invite/VJRuU3b… #tabpfn #tabularfoundationalmodels #priorlabs

TabPFN has 1.9M+ downloads! Want to join? Our first-ever webinar is Oct 13. 🔷Live demos 🔷Workflow best practices 🔷Tips to unlock TabPFN Speakers: @FrankRHutter (Co-founder Prior Labs) & Carl McBride Ellis (Kaggle Grandmaster) 👉Reserve your spot: luma.com/5c87ira0?utm_c…

Join our mission to strengthen AI research in Europe 🇪🇺 We are looking for several ML Research Engineers and Scientists to work on OpenEuroLLM at the ELLIS Institute Tübingen. If you're passionate about large-scale model training, multilingual evaluation and want to contribute to cutting-edge open-source AI, we'd love to hear from you! 📍Based in Tübingen - a vibrant hub for machine learning research with a strong collaborative ecosystem 📚 Focus: LLMs, distributed training, evaluation, AutoML 🇪🇺 Part of OpenEuroLLM - a consortium of 20 leading European research institutions, companies and EuroHPC centres to build the next-generation of open-source language models awarded many millions GPU hours on most-recent Euro HPC clusters. What we offer: 🔒 The unique chance to develop the next generation of fully open European LLMs 🤗 A chance to be in the heart of the Cyber Valley and work at the ELLIS Institute with world known scientists headed by Bernhard Schölkopf ️☕ The best coffee you can think of See more in the job post description Job Post: Building Open Source LLM for Europe: institute-tue.ellis.eu/en/jobs/openeu… Apply here: ellis-openeurollm-apply.tue.mpg.de/registration/o… #AIJob#LLM #OpenSourceAI #ResearchEngineering #OpenEUROLLM #Hiring #MachineLearning #ArtificialIntelligence

We’re excited to welcome exceptional talent from the #ELLIS network and beyond to @prior_labs! We’re building a diverse world-class team and warmly welcome applications from women and underrepresented groups in tech. Please share with your network 😊

@pirwot Too many :-) I would encourage you to follow @prior_labs and announcements / use cases in our Discord: discord.com/channels/12855…

@FrankRHutter That's awesome to hear! I'm curious, have you personally tested TabPFN v2 on any specific datasets? Any standout results or insights you can share?

It's great to have this new open-source benchmark for tabular data 🚀 It's really comprehensive, created and maintained by serious open source contributors from various groups, and I expect it to quickly become the new standard benchmark. I'm super excited that progress in the tabular space is exploding! It's also interesting to see that, for the datasets in TabPFN v2's constraints (<10k data points, <500 features), half a year after its release, TabPFN v2 still strongly outperforms all other methods!

@quantmaxxer It’s a benchmark for models, not for AutoML systems. That could be a future extension.

@FrankRHutter Sorry for laziness (paper is on my to-do list) but do they cover TabPFN with the workarounds for larger datasets? :)

@MarvinTBaumann xLSTM does well on sequential data (like time series), but it’s not as well suited to set valued data (like tabular). All the current SOTA ICL methods for tabular are transformer-based.

@FrankRHutter What about xLSTM?