NVIDIA launches Kumo Tabular, open-weights foundation models for tabular data
jure · x · 2026-09-29
Jure Leskovec announces NVIDIA Kumo Tabular, a new family of foundation models for tabular data.
- Establishes a new Pareto frontier across the entire accuracy–inference-time tradeoff
- Fully open: open weights, open-source software, and a permissive license for commercial use
- Available on Hugging Face and GitHub
Related event: NVIDIA Open-Sources Kumo Tabular Foundation Models for Tabular Data(3 posts)→
More from Models
- All models get ultrafast: ~300 tps, 8x faster but 6x more expensive — cneuralnetwork · 2026-09-30
- Ultrafast pricing sparks debate: 8x speed for 6x price needs a new plan — kevinkern · 2026-09-30
- OpenAI's new Pro 500 tier spotted, alongside GPT-Sol 6.1 — omarsar0 · 2026-09-30
- OpenAI's Pro 500 monthly plan arrives: who's paying up? — rudrank · 2026-09-30
- GPT-6.1 Sol matches GPT-6 Astra at ~1/5 cost, ultrafast version targets 300 tok/s — haider1 · 2026-09-30
- GPT-6.1 Sol Ultrafast Goes Live: Up to 300 token/s, 8x Faster at 6x the Price — kimmonismus · 2026-09-30