Category ranking · 7 models · checked 2026-10-13

Time series

Ranked by GIFT-Eval MASE, geometric mean relative to seasonal naive (lower is better) (lower is better). Source: Computed by The Local Bench from SalesforceAIResearch/gift-eval results/*/all_results.csv (97 configurations).

Granite TTM r21.0
Chronos-Bolt small0.8
Chronos-Bolt base0.8
TimesFM 2.0 500M0.8
Moirai 2.0 small0.7
TimesFM 2.50.7
Chronos-20.7

GIFT-Eval MASE, geometric mean relative to seasonal naive (lower is better). Shorter is better. Values are exactly as published; the table below links each one.

GIFT-Eval MASE, geometric mean relative to seasonal naive (lower is better)

7 with a published score, 0 not yet measured.

#	Model	Parameters	Score	Source	Licence
1	Chronos-2 Amazon	119 M	0.698Sourced	Source Computed: MASE 0.698, CRPS 0.485 vs seasonal naive; no test-data leakage per config.json.	Apache-2.0
2	TimesFM 2.5 Google	231 M	0.705Sourced	Source Computed: MASE 0.705, CRPS 0.490; no leakage per config.json.	Apache-2.0
3	Moirai 2.0 small Salesforce	11 M	0.728Sourced	Source Computed: MASE 0.728, CRPS 0.516; no leakage per config.json.	CC-BY-NC-4.0 Check terms before commercial use
4	TimesFM 2.0 500M Google	499 M	0.758Sourced	Source Computed: MASE 0.758, CRPS 0.550. config.json flags test-data leakage: treat as optimistic.	Apache-2.0
5	Chronos-Bolt base Amazon	205 M	0.808Sourced	Source Computed: MASE 0.808, CRPS 0.574. config.json flags test-data leakage.	Apache-2.0
6	Chronos-Bolt small Amazon	48 M	0.822Sourced	Source Computed from chronos_bolt_small results: MASE 0.822, CRPS 0.577. config.json flags test-data leakage.	Apache-2.0
7	Granite TTM r2 IBM	1 M	1.02Sourced	Source Computed from TTM-R2-Pretrained: MASE 1.020, CRPS 0.873. config.json flags test-data leakage.	Apache-2.0

How to read this: one public benchmark per category, so you can compare like with like. We copy the published number and link it; we do not re-run it, average different benchmarks or fill gaps. Vendor-reported numbers use each vendor's own settings, so treat small gaps as ties. “Not yet measured” means we found no comparable public score, not that the model is weak.

Every model was checked against the Hugging Face API on 2026-10-13: it exists, is public, and has the parameter count and licence shown. Other categories: Math and reasoning · Coding · Vision-language · OCR and documents · Speech to text · Text to speech · Image generation · Embeddings and rerankers · Small fast 'System 1' · Biomedical · Tool use and agents · Multilingual and translation · Robotics (VLA).

Machine-readable: /api/v1/categories/time-series.json