IBM Releases Granite Time Series Ensemble Forecasting Models
Understand the story
On October 22, 2026, IBM released the Granite Timeseries Ensemble R1 model on Hugging Face. The model combines four Granite pretrained time series models of different architectures into a single ensemble forecasting system, involving PatchTST-FM-r1, PatchTST-FM-r2, FlowState-r1.1, and TTM-r3. According to the Hugging Face Blog (firsthand source), the ensemble method completes the combination without retraining the models or fitting weights. The model outputs include point forecasts and 0.1–0.9 quantile forecasts, and can be used for time series forecasting tasks. This is currently the only report on the event, with no follow-up developments or evaluation results disclosed.
AI overview of reporting · Updated 2 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Hugging Face BlogEnsemble forecasting with IBM Granite time-series models
IBM has released Granite Timeseries Ensemble R1 on Hugging Face, combining four Granite pre-trained time-series models of different architectures (PatchTST-FM-r1/r2, FlowState-r1.1, TTM-r3) into an ensemble forecaster that requires no retraining or weight fitting and outputs point forecasts and 0.1–0.9 quantile forecasts.
Story heat trend
Not enough continuous observations to show a trend yet.