IBM open-sources Granite time-series ensemble forecasting model
What happened
On 22 October 2026, IBM published Granite Timeseries Ensemble on the Hugging Face blog, combining four pretrained Granite time-series models — PatchTST-FM-r1, PatchTST-FM-r2, FlowState-r1.1 and TTM-r3 — into a unified forecasting interface. Reports say the ensemble requires neither retraining nor fitting ensemble weights. So far the model has been released on Hugging Face; applications and evaluations have not yet been reported.
Written by AI from the reports · updated 1 h ago
Timeline
Follow the reports to see the event from each side.
- Hugging Face BlogEnsemble forecasting with IBM Granite time-series models
Four open Granite models, one extensible forecasting interface, and less guesswork about which model to use. One recipe combines four complementary Granite architectures and returns median and quantile forecasts, with no retraining and no fitted ensemble weights.
Heat of this event
Not enough continuous observations yet to draw a trend.