GIFT-Eval (MASE) gift-eval-mase Leaderboard
GIFT-Eval (Salesforce) general time-series forecasting benchmark — 23 datasets, 144K series, 7 domains, 10 frequencies. Aggregate MASE (Mean Absolute Scaled Error) across all datasets; lower is better. The certified public board of statistical, deep-learning and foundation forecasting models. · Metric: MASE (lower is better)
| # | Model | MASE | Paper |
|---|---|---|---|
| 1 | TurkForecast-FM-Chronos2-LoRA-v1 | 0.66 | link |
| 2 | RAES-Conductance-Ensemble | 0.66 | link |
| 3 | Taichu-TimeSeries-Agent | 0.67 | link |
| 4 | RAES-Conductance-Ensemble-V | 0.67 | link |
| 5 | STRIDE (+Chronos-2) | 0.67 | link |
| 6 | STRIDE (+Timer-S1) | 0.67 | link |
| 7 | TimeRouter | 0.68 | link |
| 8 | Toto-2.0-FnF | 0.68 | link |
| 9 | TSOrchestra | 0.68 | link |
| 10 | Cobra-Agent | 0.68 | link |
| 11 | MoiraiAgent-leaking | 0.68 | link |
| 12 | Toto-2.0-2.5B-FT | 0.68 | link |
| 13 | metis-autocast | 0.68 | link |
| 14 | Prism | 0.68 | link |
| 15 | DeOSAlphaTimeGPTPredictor-2025 | 0.68 | link |
| 16 | Falcon-X | 0.69 | link |
| 17 | MoiraiAgent | 0.69 | link |
| 18 | Credence | 0.69 | link |
| 19 | ZooCast-Top1 | 0.69 | — |
| 20 | Timer-S1 | 0.69 | link |
| 21 | Toto-2.0-2.5B | 0.70 | link |
| 22 | Migas-1.0 | 0.70 | link |
| 23 | Chronos-2 | 0.70 | link |
| 24 | Samay | 0.70 | — |
| 25 | TSOrchestra-test | 0.70 | link |
| 26 | Synapse | 0.70 | link |
| 27 | Toto-2.0-1B | 0.70 | link |
| 28 | FlowState-r1.1 | 0.70 | link |
| 29 | Granite-FlowState-r1.1 | 0.70 | link |
| 30 | Xihe-ultra | 0.70 | link |