IBM Granite Timeseries Ensemble combines four models in one forecasting interface
Ensemble forecasting with IBM Granite time-series models
- Who
- Hugging Face
- What
- introduced Granite Timeseries Ensemble, a four-model forecasting recipe
- Quote
Granite Timeseries Ensemble reduces that choice to a sensible default.
What happened
IBM introduces Granite Timeseries Ensemble, a permissively licensed recipe combining four pretrained Granite models to produce median and quantile forecasts without retraining or fitted ensemble weights. It offers default equal-influence probability pooling and an adaptive IQR alternative. Across 97 GIFT-Eval configurations, the default achieves aggregate MASE of 0.683 versus 0.685 for the strongest individual member, while PatchTST-FM-r2 retains the lowest aggregate CRPS/MWQL. Loading four checkpoints increases inference time and memory use.
Why it matters
The benchmark comparison clarifies the ensemble’s tradeoff between slightly better aggregate point forecasts and increased inference costs, while showing why probabilistic accuracy can favor an individual member.
Read next
See the full storyIBM Introduces Granite Timeseries Ensemble to Combine Four Forecasting ModelsThe story so far
IBM introduced Granite Timeseries Ensemble, a permissively licensed recipe that combines four pretrained Granite models in one interface to produce median and quantile forecasts without retraining or fitting ensemble weights. In the Hugging Face blog report’s evaluation across 97 GIFT-Eval configurations, the default ensemble achieved aggregate MASE of 0.683, compared with 0.685 for the strongest individual model. PatchTST-FM-r2 retained the lowest aggregate CRPS/MWQL, so the ensemble did not lead on every reported metric. The recipe defaults to equal-influence probability pooling and also offers an adaptive IQR alternative. It stores an ensemble configuration rather than merged model weights and downloads all four member checkpoints at runtime, adding time to the first run. Running four models also increases inference time and memory use. The recipe and its default Granite members are permissively licensed.
Keep up with AI, one email a day
Model releases, price changes and compute developments, with original sources. Free, confirmed by email, and easy to unsubscribe. Read a past issue · 2026-10-22