GIFT-Eval
An authoritative evaluation benchmark for time-series forecasting, covering combinations of multiple datasets, frequencies, and forecasting horizon configurations. It uses MASE as a unified evaluation metric and supports cross-dataset comparisons.
Timeline (last 90 days)
On the GIFT-Eval benchmark, the best orchestrator reduces MASE by less than 6% relative to the best-performing single TSFM.
The 209M-parameter Timestral leads in the zero-shot model category on three benchmarks: GIFT-Eval, TIME and fev-bench.
GIFT-Eval is a zero-shot time series evaluation benchmark proposed by researchers from Salesforce and other institutions, covering real datasets in multiple domains such as energy, transportation, retail, weather and finance, and distinguishing between univariate and multivariate tasks.
Mistral says the 209M version of Timestral ranks first among zero-shot pretrained models on GIFT-Eval/TIME/fev-bench.
Timestral-28M scores 0.678 on GIFT-Eval MASE and 0.638 on TIME MASE.
Timestral-209M scores 0.658 MASE on GIFT-Eval, versus 0.661 for t0-gamma, 0.667 for TimesFM-3, and 0.698 for Chronos-2.
TW3Cast ranks 3rd on the GIFT-Eval time series forecasting benchmark (out of 130 entries, ranked by average MASE), with the ranking snapshot dated 2026-09-14.
TW3Cast faces 97 combinations of dataset, frequency, and forecast horizon configurations on GIFT-Eval, and the routing table selects one of four modes for each configuration.
The two GIFT-Eval entries ranked above TW3Cast belong to the agentic category of the leaderboard.
All Facts (7)
On the GIFT-Eval benchmark, the best orchestrator reduces MASE by less than 6% relative to the best-performing single TSFM.
50%UnverifiedThe 209M-parameter Timestral leads in the zero-shot model category on three benchmarks: GIFT-Eval, TIME and fev-bench.
50%UnverifiedGIFT-Eval is a zero-shot time series evaluation benchmark proposed by researchers from Salesforce and other institutions, covering real datasets in multiple domains such as energy, transportation, retail, weather and finance, and distinguishing between univariate and multivariate tasks.
50%UnverifiedMistral says the 209M version of Timestral ranks first among zero-shot pretrained models on GIFT-Eval/TIME/fev-bench.
50%UnverifiedTimestral-28M scores 0.678 on GIFT-Eval MASE and 0.638 on TIME MASE.
50%UnverifiedTimestral-209M scores 0.658 MASE on GIFT-Eval, versus 0.661 for t0-gamma, 0.667 for TimesFM-3, and 0.698 for Chronos-2.
50%UnverifiedTW3Cast faces 97 combinations of dataset, frequency, and forecast horizon configurations on GIFT-Eval, and the routing table selects one of four modes for each configuration.
50%