NVIDIA has launched Kumo Tabular, a new family of open foundation models for tabular data, now available on Hugging Face. These models, part of the NVIDIA Kumo Structured collection, can predict labels for new rows in a single forward pass without requiring training or tuning. Kumo Tabular, which was pretrained on artificial data and comes in three sizes, has demonstrated top performance on several benchmarks, outperforming existing models and even GPT-4o on table understanding tasks. The models are released under the permissive OpenMDW-1.1 license, allowing for commercial use. AI
IMPACT Sets a new standard for tabular data prediction, potentially shifting enterprise ML away from traditional gradient-boosted trees.
RANK_REASON NVIDIA released a new family of tabular foundation models with benchmark performance claims.
Read on Hugging Face Trending Models →
- BeyondArena
- GitHub
- Hugging Face
- Kumo Tabular
- Nvidia
- NVIDIA Kumo Structured
- OpenMDW-1.1
- ScoringBench
- TabArena
- TabICL
- TabPFN
- GPT-4o
- structured-data-models
AI-generated summary · Google Gemini · from 4 sources. How we write summaries →