NVIDIA debuts open AI model for predicting from tabular data
NVIDIA has launched Kumo Tabular, an open-source AI model designed for making predictions directly from structured tables, like customer records or transaction logs. Unlike previous models that needed custom tuning, it predicts class labels or values for new rows in one step, with no retraining or manual data prep. It reportedly tops four public benchmarks, comes in three sizes, and is free for commercial use.
- Works with both classification and regression tasks
- No feature engineering or training needed
- Pretrained only on synthetic tables
- Available in three sizes: 28M to 215M parameters
- Released for commercial use under OpenMDW-1.1
Sources covering this
In this story
More in AI
OpenAI launches Dots AI agents amid delayed model update
OpenAI has released Dots, its new always-on AI assistant designed to autonomously handle tasks for users.
OpenAI rolls out Dots, always-on AI agents for work tasks
OpenAI has launched Dots, always-available AI assistants designed to handle projects, tasks, and reminders across apps like Slack,…
Nvidia releases Open Agent Safety Platform to contain rogue AI
Nvidia has launched the Open Agent Safety Platform, aimed at containing AI agents that attempt to circumvent their boundaries.
OpenAI cancels GPT-6.1 Astra over safety failures
OpenAI has called off the upcoming release of its GPT-6.1 Astra model after internal tests flagged safety problems.