Prior Labs ships TabPFN-3.5, SOTA tabular foundation model leading both TabArena and BeyondArena
tuanacelik · reddit · 2026-09-16
Prior Labs released TabPFN-3.5, topping both TabArena and BeyondArena and hitting SOTA at 1M rows and up to 20k features. It ships in three variants: Fast (alpha, 6x faster than base), Thinking (API-based compute-for-accuracy tradeoff, +20 Elo on BeyondArena and +44 on TabArena over base), and Plus. On BeyondArena it leads text-rich, high-cardinality, and high-dimensional data by 250+ Elo over the strongest prior baseline and 150+ Elo over the previous overall leader.
Related event: Prior Labs Releases TabPFN-3.5, Topping Table Benchmarks(3 posts)→
More from Models
- Google launches Gemini 3.8 Live and 3.8 Live Extended Thinking with 97-language real-time voice — GoogleAI · 2026-09-16
- Google rolls out Gemini 3.8 Live audio models to consumers, developers and enterprises — GoogleAI · 2026-09-16
- DeepMind demos Gemini 3.8 Live as a coding tutor that narrates its reasoning — GoogleDeepMind · 2026-09-16
- Google DeepMind launches Gemini 3.8 Live and Live Extended Thinking — GoogleDeepMind · 2026-09-16
- p0 efficiency ranking: GPT-5.6 Luna tops at $23.03 per 1,000 tasks — TheMoonMidas · 2026-09-16
- OpenArt Arena launches a pro blind-test leaderboard for creative AI models — azed_ai · 2026-09-16