NVIDIA open-sources Kumo Forecast time series forecasting model
Understand the story
On October 21, 2026, NVIDIA released the multivariate time series forecasting model Kumo Forecast on the Hugging Face Blog. The model was originally named NV-Tesseract Forecasting. Its architecture is based on the MOMENT-1-large Transformer encoder with about 340 million parameters, adding a trainable forecasting head on top of it. The model combines pretrained time series representations with retrieval of historical patterns: through the DARR pattern, it retrieves historical patterns in a k-nearest-neighbor manner to enhance direct forecasting performance, while emphasizing interpretability. This is the first report on the event; at present the event only consists of this release information, with no follow-up developments reported yet.
AI overview of reporting · Updated 3 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Hugging Face BlogNVIDIA Kumo Forecast Combines Pretrained Time-Series Representations with Historical Pattern Retrieval and Interpretability
NVIDIA released Kumo Forecast (formerly NV-Tesseract Forecasting), a multivariate time-series forecasting model built on the ~340M-parameter MOMENT-1-large Transformer encoder with a trainable forecasting head, and enhanced via a DARR mode that uses k-nearest-neighbor retrieval of historical patterns to augment direct predictions.
Story heat trend
Current heat9·Comparable peak10（10/21 21:00）·Comparable change over 24 hours –
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.