GIFT-Eval
Probabilistic and point forecasting across 97 dataset configurations; MASE and CRPS normalised by Seasonal Naive
time seriesactive
Score in mean CRPS rank, Toto 2.0 report (Datadog)
Definition
- Version
- not in the notes
- Category
- time series
- Maintainer
- Salesforce
- Status
- active
- Home page
- https://huggingface.co/spaces/Salesforce/GIFT-Eval
Sources
- GIFT Eval - a Hugging Face Space by Salesforce huggingface.co https://huggingface.co/spaces/Salesforce/GIFT-Eval
- arxiv.org/pdf/2605.20119 https://arxiv.org/pdf/2605.20119
- GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation arxiv.org https://arxiv.org/abs/2410.10393
- Introducing Chronos-2: From univariate to universal forecasting amazon.science https://www.amazon.science/blog/introducing-chronos-2-from-univariate-to-universal-forecasting
- Google TimesFM-3 voegt multivariate forecasting toe datanorth.ai https://datanorth.ai/nl/nieuws/google-lanceert-timesfm-3
- Introducing Moirai 2.0 salesforce.com https://www.salesforce.com/blog/moirai-2-0/
- Moirai 2.0: When Less Is More for Time Series Forecasting arxiv.org https://arxiv.org/html/2511.11698v1
- Granite Time Series | IBM Granite ibm.com https://www.ibm.com/ae-ar/granite/docs/models/time-series/
Scores at a Glance
Score in mean CRPS rank, Toto 2.0 report (Datadog)
Lower is better for this unit.
1
Toto 2.0 2.5BDatadogzero-shot foundation-model CRPS rank, lower is betterToto 2.0 1BDatadogzero-shot foundation-model CRPS rank, lower is betterToto 2.0 313MDatadogzero-shot foundation-model CRPS rank, lower is better
Granite PatchTST-FM r1IBM, Rensselaer Polytechnic Institutezero-shot foundation-model CRPS rank, lower is better
Chronos-2Amazonzero-shot foundation-model CRPS rank, lower is better
TiRexNX-AIzero-shot foundation-model CRPS rank, lower is betterFlowState r1.1IBMzero-shot foundation-model CRPS rank, lower is better
TimesFM 2.5Googlezero-shot foundation-model CRPS rank, lower is better
Timer-S1Tsinghua University, ByteDancezero-shot foundation-model CRPS rank, lower is better
Xihe-ultraAnt Groupzero-shot foundation-model CRPS rank, lower is betterMoirai 2.0 smallSalesforcezero-shot foundation-model CRPS rank, lower is better
Toto 2.0 2.5BDatadogzero-shot foundation-model CRPS rank, lower is better20.3 mean CRPS rank
vendor-reported2Toto 2.0 1BDatadogzero-shot foundation-model CRPS rank, lower is better21.1 mean CRPS rank
vendor-reported3Toto 2.0 313MDatadogzero-shot foundation-model CRPS rank, lower is better21.4 mean CRPS rank
vendor-reported4Granite PatchTST-FM r1IBM, Rensselaer Polytechnic Institutezero-shot foundation-model CRPS rank, lower is better23.1 mean CRPS rank
vendor-reported5Chronos-2Amazonzero-shot foundation-model CRPS rank, lower is better23.5 mean CRPS rank
vendor-reported6TiRexNX-AIzero-shot foundation-model CRPS rank, lower is better23.9 mean CRPS rank
vendor-reported7FlowState r1.1IBMzero-shot foundation-model CRPS rank, lower is better25.6 mean CRPS rank
vendor-reported8TimesFM 2.5Googlezero-shot foundation-model CRPS rank, lower is better25.8 mean CRPS rank
vendor-reported9Timer-S1Tsinghua University, ByteDancezero-shot foundation-model CRPS rank, lower is better26.5 mean CRPS rank
vendor-reported10Xihe-ultraAnt Groupzero-shot foundation-model CRPS rank, lower is better29.5 mean CRPS rank
vendor-reported11Moirai 2.0 smallSalesforcezero-shot foundation-model CRPS rank, lower is better36.5 mean CRPS rank
vendor-reported010203040
Score in MASE (geo-mean, normalised to Seasonal Naive), Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations, zero-shot
Lower is better for this unit.
1
Lingjiang2.0Alibaba (Qwen)TimesFM 3.0Google
EXAONE-ForecastLG AI ResearchGranite PatchTST-FM r2IBM, Rensselaer Polytechnic InstituteXihe-ultraAnt GroupFlowState r1.1IBMTimesFM 2.5GoogleTiRexNX-AISundial (Timer 3.0) base-128mTsinghua University
TabPFN-TSPrior Labs
Lingjiang2.0Alibaba (Qwen)0.666 MASE (geo-mean, normalised to Seasonal Naive)
independent2TimesFM 3.0Google0.667 MASE (geo-mean, normalised to Seasonal Naive)
independent3EXAONE-ForecastLG AI Research0.673 MASE (geo-mean, normalised to Seasonal Naive)
independent4Granite PatchTST-FM r2IBM, Rensselaer Polytechnic Institute0.685 MASE (geo-mean, normalised to Seasonal Naive)
independent5Xihe-ultraAnt Group0.701 MASE (geo-mean, normalised to Seasonal Naive)
independent6FlowState r1.1IBM0.701 MASE (geo-mean, normalised to Seasonal Naive)
independent7TimesFM 2.5Google0.705 MASE (geo-mean, normalised to Seasonal Naive)
independent8TiRexNX-AI0.716 MASE (geo-mean, normalised to Seasonal Naive)
independent9Sundial (Timer 3.0) base-128mTsinghua University0.75 MASE (geo-mean, normalised to Seasonal Naive)
independent10TabPFN-TSPrior Labs0.771 MASE (geo-mean, normalised to Seasonal Naive)
independent00.511.5
Score in CRPS (geo-mean, normalised to Seasonal Naive), Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations, zero-shot; rank is average CRPS rank
Lower is better for this unit.
1TimesFM 3.0GoogleLingjiang2.0Alibaba (Qwen)EXAONE-ForecastLG AI ResearchGranite PatchTST-FM r2IBM, Rensselaer Polytechnic InstituteFlowState r1.1IBMXihe-ultraAnt GroupTiRexNX-AITimesFM 2.5GoogleTabPFN-TSPrior LabsSundial (Timer 3.0) base-128mTsinghua University
TimesFM 3.0Google0.456 CRPS (geo-mean, normalised to Seasonal Naive)
independent2Lingjiang2.0Alibaba (Qwen)0.457 CRPS (geo-mean, normalised to Seasonal Naive)
independent3EXAONE-ForecastLG AI Research0.46 CRPS (geo-mean, normalised to Seasonal Naive)
independent4Granite PatchTST-FM r2IBM, Rensselaer Polytechnic Institute0.467 CRPS (geo-mean, normalised to Seasonal Naive)
independent5FlowState r1.1IBM0.487 CRPS (geo-mean, normalised to Seasonal Naive)
independent6Xihe-ultraAnt Group0.488 CRPS (geo-mean, normalised to Seasonal Naive)
independent7TiRexNX-AI0.488 CRPS (geo-mean, normalised to Seasonal Naive)
independent8TimesFM 2.5Google0.49 CRPS (geo-mean, normalised to Seasonal Naive)
independent9TabPFN-TSPrior Labs0.544 CRPS (geo-mean, normalised to Seasonal Naive)
independent10Sundial (Timer 3.0) base-128mTsinghua University0.559 CRPS (geo-mean, normalised to Seasonal Naive)
independent00.250.50.751
Score in MASE (geo-mean, normalised to Seasonal Naive), Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations, pretrained
Lower is better for this unit.
1
ChakraTSYHat Labs
QiYao-MECNU, Huawei (Huawei Cloud)TiRex-2NX-AITimer-S1Tsinghua University, ByteDanceToto 2.0 2.5BDatadogChronos-2AmazonMoirai 2.0 smallSalesforceGranite TinyTimeMixer (TTM) r2IBM
Lag-LlamaMorgan Stanley, ServiceNow (ServiceNow Research, Montreal), Universite de Montreal, Mila
ChakraTSYHat Labs0.67 MASE (geo-mean, normalised to Seasonal Naive)
independent2QiYao-MECNU, Huawei (Huawei Cloud)0.671 MASE (geo-mean, normalised to Seasonal Naive)
independent3TiRex-2NX-AI0.678 MASE (geo-mean, normalised to Seasonal Naive)
independent4Timer-S1Tsinghua University, ByteDance0.693 MASE (geo-mean, normalised to Seasonal Naive)
independent5Toto 2.0 2.5BDatadog0.696 MASE (geo-mean, normalised to Seasonal Naive)
independent6Chronos-2Amazon0.698 MASE (geo-mean, normalised to Seasonal Naive)
independent7Moirai 2.0 smallSalesforce0.728 MASE (geo-mean, normalised to Seasonal Naive)
independent8Granite TinyTimeMixer (TTM) r2IBM1.02 MASE (geo-mean, normalised to Seasonal Naive)
independent9Lag-LlamaMorgan Stanley, ServiceNow (ServiceNow Research, Montreal), Universite de Montreal, Mila1.228 MASE (geo-mean, normalised to Seasonal Naive)
independent00.511.5
Score in CRPS (geo-mean, normalised to Seasonal Naive), Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations, pretrained; rank is average CRPS rank
Lower is better for this unit.
1QiYao-MECNU, Huawei (Huawei Cloud)ChakraTSYHat LabsTiRex-2NX-AIToto 2.0 2.5BDatadogChronos-2AmazonTimer-S1Tsinghua University, ByteDanceMoirai 2.0 smallSalesforce
QiYao-MECNU, Huawei (Huawei Cloud)0.461 CRPS (geo-mean, normalised to Seasonal Naive)
independent2ChakraTSYHat Labs0.461 CRPS (geo-mean, normalised to Seasonal Naive)
independent3TiRex-2NX-AI0.467 CRPS (geo-mean, normalised to Seasonal Naive)
independent4Toto 2.0 2.5BDatadog0.476 CRPS (geo-mean, normalised to Seasonal Naive)
independent5Chronos-2Amazon0.485 CRPS (geo-mean, normalised to Seasonal Naive)
independent6Timer-S1Tsinghua University, ByteDance0.485 CRPS (geo-mean, normalised to Seasonal Naive)
independent7Moirai 2.0 smallSalesforce0.516 CRPS (geo-mean, normalised to Seasonal Naive)
independent00.250.50.751
Score in CRPS (geo-mean, normalised to Seasonal Naive), Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations, pretrained
Lower is better for this unit.
1Granite TinyTimeMixer (TTM) r2IBMLag-LlamaMorgan Stanley, ServiceNow (ServiceNow Research, Montreal), Universite de Montreal, Mila
Granite TinyTimeMixer (TTM) r2IBM0.873 CRPS (geo-mean, normalised to Seasonal Naive)
independent2Lag-LlamaMorgan Stanley, ServiceNow (ServiceNow Research, Montreal), Universite de Montreal, Mila0.88 CRPS (geo-mean, normalised to Seasonal Naive)
independent00.250.50.751
Where this comes from
Each bar is one model, with its most independent recorded score. Hatched bars are vendor-reported; the word beside each bar says who ran it. Settings and dates differ between rows; the Scores table below has every result.
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
| Moirai 2.0 small | 36.5 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Xihe-ultra | 29.5 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Timer-S1 | 26.5 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| TimesFM 2.5 | 25.8 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| FlowState r1.1 | 25.6 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| TiRex | 23.9 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Chronos-2 | 23.5 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Granite PatchTST-FM r1 | 23.1 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Toto 2.0 313M | 21.4 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Toto 2.0 1B | 21.1 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Toto 2.0 2.5B | 20.3 mean CRPS rank | vendor-reported | zero-shot foundation-model CRPS rank, lower is better | Toto 2.0 report (Datadog) | Jun 2026 | arxiv.org |
| Lag-Llama | 1.228 MASE (geo-mean, normalised to Seasonal Naive) | independentsaw test data | pretrained, test leakage Yes | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Auto_Arima | 1.074 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | statistical | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Granite TinyTimeMixer (TTM) r2 | 1.02 MASE (geo-mean, normalised to Seasonal Naive) | independentsaw test data | pretrained, test leakage Yes | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Seasonal_Naive | 1 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | baseline; rank is average CRPS rank | rank 138 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Seasonal_Naive | 1 MASE (geo-mean, normalised to Seasonal Naive) | independent | baseline | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Auto_Arima | 0.912 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | statistical; rank is average CRPS rank | rank 131 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Lag-Llama | 0.88 CRPS (geo-mean, normalised to Seasonal Naive) | independentsaw test data | pretrained, test leakage Yes; rank is average CRPS rank | rank 136 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Granite TinyTimeMixer (TTM) r2 | 0.873 CRPS (geo-mean, normalised to Seasonal Naive) | independentsaw test data | pretrained, test leakage Yes; rank is average CRPS rank | rank 140 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TabPFN-TS | 0.771 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Sundial (Timer 3.0) base-128m | 0.75 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Moirai 2.0 small | 0.728 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Moirai 2.0 small | 0.728 MASE (normalised) | vendor-reported | – | rank 5 of 30, 30 foundation models | Nov 2025 | arxiv.org |
| TiRex | 0.716 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TimesFM 2.5 | 0.705 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Xihe-ultra | 0.701 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| FlowState r1.1 | 0.701 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Chronos-2 | 0.698 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Toto 2.0 2.5B | 0.696 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Timer-S1 | 0.693 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Granite PatchTST-FM r2 | 0.685 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TiRex-2 | 0.678 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| EXAONE-Forecast | 0.673 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| QiYao-M | 0.671 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| ChakraTS | 0.67 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TimesFM 3.0 | 0.667 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Lingjiang2.0 | 0.666 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| STRIDE_w_Synapse | 0.625 MASE (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | agentic | Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Sundial (Timer 3.0) base-128m | 0.559 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 106 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TabPFN-TS | 0.544 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 93 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Moirai 2.0 small | 0.516 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 81 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Moirai 2.0 small | 0.516 CRPS (normalised) | vendor-reported | – | rank 6 of 30, 30 foundation models | Nov 2025 | arxiv.org |
| TimesFM 2.5 | 0.49 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 66 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Xihe-ultra | 0.488 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 73 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TiRex | 0.488 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 62 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| FlowState r1.1 | 0.487 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 64 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Chronos-2 | 0.485 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 57 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Timer-S1 | 0.485 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 68 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Toto 2.0 2.5B | 0.476 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 50 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Granite PatchTST-FM r2 | 0.467 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 36 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TiRex-2 | 0.467 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 45 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| QiYao-M | 0.461 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 34 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| ChakraTS | 0.461 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | pretrained; rank is average CRPS rank | rank 26 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| EXAONE-Forecast | 0.46 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 22 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Lingjiang2.0 | 0.457 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 24 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| TimesFM 3.0 | 0.456 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | zero-shot; rank is average CRPS rank | rank 14 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| STRIDE_w_Synapse | 0.423 CRPS (geo-mean, normalised to Seasonal Naive) | independentno test-data leakage | agentic; rank is average CRPS rank | rank 1 of 143, Leaderboard rebuilt from the published raw results, 143 entries covering all 97 configurations | 9 Oct 2026 | huggingface.co |
| Chronos-2 | – | vendor-reported | Claims first among pretrained models, beating TimesFM-2.5 and TiRex | rank 1, – | 20 Oct 2025 | amazon.science |
| TimesFM 3.0 | – | vendor-reported | Google claims best pretrained FM; no numbers given | rank 1, – | 31 Aug 2026 | datanorth.ai |
| Granite FlowState | – | vendor-reported | IBM Granite docs | rank 3, – | 2 Oct 2025 | ibm.com |
| Moirai 2.0 | – | vendor-reported | arXiv paper ranking among foundation models | rank 5 of 37, – | 12 Nov 2025 | arxiv.org |
| Moirai 2.0 | – MASE | vendor-reported | Salesforce blog: #1 by MASE among non-leaking FMs | rank 1, – | 8 Aug 2025 | salesforce.com |
| Sundial (Timer 3.0) base-128m | – MASE rank | vendor-reported | MASE | rank 1, – | May 2025 | huggingface.co |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "GIFT-Eval", https://aurion.semai.id/benchmarks/gift-eval/, records of October 2026.