Prior Labs model beats the 2015 Otto Kaggle winner with default settings
TabPFN-3.5 scored 0.375 on the Otto competition's private leaderboard using raw data and default settings, beating the 2015 winning entry.
Prior Labs has released TabPFN-3.5, a new version of its tabular foundation model. It predicts on a table in a single forward pass, with no per-dataset training or tuning. The company reports first place across 7 tabular benchmarks.
Open weights run locally for research, evaluation and Kaggle use. Production deployment requires Prior Labs' API or a commercial license.
The Otto score
The Otto Group Product Classification Challenge ran on Kaggle in 2015, drawing 3,505 teams competing for $10,000. Entrants sorted products into 9 categories using 93 obfuscated count features, scored with multi-class log loss, where lower is better.
The winning entry came from Gilberto Titericz and Stanislav Semenov, both former world #1 Kaggle grandmasters. It was a multi-layer stack of 36 models built on hand-crafted features.
Nick Erickson, co-creator of AutoGluon and an AI researcher at Prior Labs, has chased that score for years. He says AutoGluon placed 23rd in its 2020 paper, AutoGluon 1.0 reached rank 14 in 2023, and AutoGluon 1.6 reached rank 9 in August 2026. The last stretch was hardest: going from rank 50 to rank 10 cut log loss from 0.41 to 0.40, while reaching the winning 0.382 required a further 0.018.
TabPFN-3.5 scores 0.375 on the private leaderboard. Erickson says it ran on raw data with default settings, taking about a minute on an RTX PRO 6000 GPU. The model was pretrained only on synthetic data and never saw Otto or any other Kaggle dataset. A reproducible Kaggle notebook is public.
Benchmark results
The technical report lists first place on TabArena, BeyondArena, STRABLE, MulTaBench, RelArena-α, TALENT and ScoringBench. The top entry is not always the base model: TabPFN-3.5-Thinking leads TabArena, BeyondArena, STRABLE and MulTaBench, while an internal TabPFN-Rel harness preview takes RelArena-α.
On TabArena, a living benchmark of 51 datasets, Thinking reaches 1910 Elo. The base model scores 1866, ahead of TabFM+ at 1823. Prior Labs says the base model beats AutoGluon 1.6 extreme by 130 Elo in a fifth of the time. On BeyondArena, which spans 142 datasets, TabPFN-3.5 finishes roughly 150 Elo ahead of the previous leader, though tuned and ensembled MLPs still lead on its grouped, temporal and large-data slices.
The KV cache stays about the same size as in TabPFN-3 despite roughly 4x more parameters. Cached single-row predictions match TabPFN-3 speed, but on large training sets the base model runs up to 2x slower.
More detail is in the MarkTechPost report.
