# Zero-Shot Forecasting

Mistral released Timestral, its first family of time-series foundation models, whose largest version at 209M parameters ranks first on the public GIFT-Eval, TIME and fev-bench forecasting benchmarks while using two-thirds the parameters of the previous state of the art. It produces probabilistic multivariate forecasts with past-only and known-future covariates, spans four sizes from 5M to 209M parameters, and needs no GPU: on an 8-thread server CPU, a one-day-ahead forecast from three weeks of 15-minute history takes 10 ms at 5M and 140 ms at 209M.

Mistral released Timestral, its first family of time-series foundation models, whose largest version at 209M parameters ranks first on the public GIFT-Eval, TIME and fev-bench forecasting benchmarks while using two-thirds the parameters of the previous state of the art. It produces probabilistic multivariate forecasts with past-only and known-future covariates, spans four sizes from 5M to 209M parameters, and needs no GPU: on an 8-thread server CPU, a one-day-ahead forecast from three weeks of 15-minute history takes 10 ms at 5M and 140 ms at 209M.

On 57 day-ahead test days for a university teaching building, pretrained TimesFM 3.0 with covariates reached 19.97 kWh hourly mean absolute error, about 43% below the weekly seasonal-naive baseline, without training building-specific weights.

On 57 day-ahead test days for a university teaching building, pretrained TimesFM 3.0 with covariates reached 19.97 kWh hourly mean absolute error, about 43% below the weekly seasonal-naive baseline, without training building-specific weights.

Time series foundation models now accept covariates, yet how calendar features are encoded has been treated as an implementation detail. Across four models and nine datasets shaped by weekends and holidays, seven calendar encodings were compared in a zero-shot rolling evaluation: encoding choice changed accuracy in every model, with within-model MASE gaps up to 9.9% on non-holidays and 83% on holidays, and the best encoding differed by model. SHAP attribution locates the gains around holidays, where signals cannot be inferred from target history.

Time series foundation models now accept covariates, yet how calendar features are encoded has been treated as an implementation detail. Across four models and nine datasets shaped by weekends and holidays, seven calendar encodings were compared in a zero-shot rolling evaluation: encoding choice changed accuracy in every model, with within-model MASE gaps up to 9.9% on non-holidays and 83% on holidays, and the best encoding differed by model. SHAP attribution locates the gains around holidays, where signals cannot be inferred from target history.