GIFT-Eval (General Time Series Forecasting Model Evaluation) is a probabilistic forecasting benchmark from Salesforce that evaluates time series foundation models across 23 diverse dataset groups spanning 7 domains and 10 frequencies. Models are ranked by Average Rank and Average Weighted Quantile Loss.

What does Average Rank mean in GIFT-Eval?

Average Rank is the mean placement of a model across every benchmark slice. A lower value means the model consistently finishes near the top across heterogeneous datasets rather than dominating only one slice.

When should I use GIFT-Eval instead of FEV Bench?

Use GIFT-Eval when you care about probabilistic forecast quality (calibrated uncertainty) and robustness across many different data domains, rather than just point-forecast accuracy on a single leaderboard.

Does GIFT-Eval test multivariate forecasting?

GIFT-Eval includes both univariate and multivariate slices. The leaderboard on TSFM.ai aggregates both variate types from the upstream grouped-by-univariate file, and you can filter per-slice rankings to inspect multivariate-only performance.

What is the 'test leak' column?

Some GIFT-Eval submissions are from models whose training data overlaps with the evaluation datasets. A 'Yes' in the test-leak column means the authors disclosed partial or full pretraining-data overlap; use the filter to compare only models with no known test-data leakage.

What model types appear on the leaderboard?

GIFT-Eval ranks pretrained foundation models, zero-shot models, fine-tuned models, agentic systems, classical deep-learning architectures (PatchTST, iTransformer, TFT, TiDE, N-BEATS), and statistical baselines (ARIMA, ETS, Theta). The page shows every type so you can benchmark a TSFM against classical prior art.

How often is the leaderboard refreshed?

TSFM.ai refetches the upstream Salesforce GIFT-Eval results every 12 hours via Next.js ISR. The 'last refreshed' timestamp at the bottom of the leaderboard reflects the most recent successful refresh.

Live·94 models · 94 evaluated·Auto-refreshed from the official leaderboard every 12 hours

GIFT-Eval Leaderboard

GIFT-Eval expands the question beyond raw point accuracy. It looks at how models perform across many datasets, frequencies, and forecasting settings, making it a strong benchmark for teams that care about robustness rather than a single flattering leaderboard slice. This page mirrors the official Salesforce GIFT-Eval leaderboard with search, filters, and per-slice rankings — covering every submitted model, not just the top few — and links each result back to a hosted endpoint on TSFM.ai where available.

Cobra-Agent ranks first with an average MASE rank of 14.63 across the GIFT-Eval dataset surface.

What this benchmark answers

Which models stay strong across heterogeneous datasets and probabilistic settings?

Methodology

Models are scored on grouped benchmark slices and ranked by average rank, with Weighted Quantile Loss providing a secondary read on probabilistic accuracy.

Showing 94 of 94 models · 37 hosted on TSFM.ai · Sorted by overall MASE rank (lower is better)

#	Model	Type	MASE rank	CRPS rank	MASE	CRPS	Leak	Repl. code	Hosted
1	Cobra-AgentHF Dalpha AI	Agentic	14.63	14.00	0.68	0.46	No	No	—
2	Toto-2.0-FnFHFCode Datadog	Agentic	16.45	15.80	0.68	0.46	No	Yes	—
3	RAES-Conductance-EnsembleHFCode California State University Northridge	Agentic	16.62	16.65	0.66	0.46	Yes	Yes	—
4	Taichu-TimeSeries-AgentHFCode zidongtaichu	Agentic	17.58	18.44	0.67	0.46	No	No	—
5	Toto-2.0-2.5B-FTHFCode Datadog	Fine-tuned	18.29	17.04	0.68	0.46	No	Yes	—
6	TSOrchestraHFCode Melady Lab @ USC	Agentic	18.41	16.98	0.68	0.47	No	No	—
7	DeOSAlphaTimeGPTPredictor-2025HF vencortex®	Zero-shot	19.03	17.51	0.68	0.47	No	No	—
8	MoiraiAgent-leakingHFCode Salesforce AI Research	Agentic	20.18	22.96	0.68	0.47	Yes	Yes	—
9	RAES-Conductance-Ensemble-VHFCode SETI	Agentic	20.34	20.48	0.67	0.46	No	Yes	—
10	MoiraiAgentHFCode Salesforce AI Research	Agentic	21.90	23.86	0.69	0.48	No	Yes	—
11	CredenceHFCode ContinualIST	Agentic	22.02	19.64	0.69	0.47	No	Yes	—
12	Samay Kairosity	Agentic	24.41	21.56	0.70	0.48	No	No	—
13	Toto-2.0-2.5BHFCode Datadog	Pretrained	25.40	24.82	0.70	0.48	No	Yes	—
14	ZooCast-Top1Code	Agentic	25.46	26.02	0.69	0.48	No	No	—
15	Migas-1.0HF Synthefy	Pretrained	25.67	65.32	0.70	0.60	No	No	—
16	STRIDE (+Chronos-2)HF Google Cloud AI Research	Pretrained	25.75	21.27	0.67	0.45	No	No	—
17	SynapseHF Google Cloud AI Research	Agentic	26.01	22.43	0.70	0.48	No	No	—
18	TimeCopilotHFCode	Agentic	26.01	24.07	0.71	0.48	No	Yes	—
19	STRIDE (+Timer-S1)HF Google Cloud AI Research	Pretrained	26.41	24.69	0.67	0.46	No	No	—
20	Toto-2.0-1BHFCode Datadog	Pretrained	26.70	25.71	0.70	0.48	No	Yes	View model
21	Toto-2.0-313mHFCode Datadog	Pretrained	27.68	26.14	0.70	0.48	No	Yes	View model
22	Chronos-2HFCode AWS	Pretrained	28.33	28.67	0.70	0.49	No	Yes	View model
23	Prism Birla AI Labs	Agentic	29.68	23.18	0.70	0.48	No	No	—
24	TurkForecast-FM-Chronos2-LoRA-v1HFCode TurkForecast (Mert Karatay)	Fine-tuned	30.09	29.49	0.66	0.42	No	Yes	—
25	TSOrchestra-testHF Melady Lab @ USC	Fine-tuned	30.18	21.79	0.70	0.47	Yes	No	—
26	PatchTST-FM-r1HFCode IBM TSFM & Rensselaer Polytechnic Institute	Zero-shot	31.36	27.95	0.71	0.48	No	Yes	View model
27	Granite-FlowState-r1.1HFCode IBM TSFM	Zero-shot	31.38	31.90	0.70	0.49	No	Yes	—
28	Timer-S1HFCode Tsinghua & ByteDance	Pretrained	31.56	31.94	0.69	0.49	No	Yes	View model
29	TimesFM-2.5HFCode Google Research	Zero-shot	32.14	31.04	0.71	0.49	No	Yes	View model
30	FlowState-r1.1HFCode IBM TSFM	Zero-shot	32.22	30.72	0.70	0.49	No	Yes	View model
31	LongSeer-v1.0 LongShine AI Research	Zero-shot	33.41	33.61	0.71	0.49	No	No	—
32	TiRexHFCode NXAI	Zero-shot	35.26	29.28	0.72	0.49	No	Yes	View model
33	Falcon-XHFCode ant-intl	Pretrained	35.27	34.80	0.69	0.49	No	Yes	—
34	Granite-PatchTST-FM-r1HFCode IBM TSFM & Rensselaer Polytechnic Institute	Zero-shot	36.06	31.03	0.72	0.49	No	Yes	View model
35	Reverso MIT	Zero-shot	36.13	67.31	0.71	0.61	No	No	—
36	Toto-2.0-22mHFCode Datadog	Pretrained	36.39	32.22	0.72	0.50	No	Yes	View model
37	Xihe-ultraHF Ant	Zero-shot	37.19	35.21	0.70	0.49	No	No	—
38	Chronos-2-SynthHFCode AWS	Zero-shot	37.47	36.30	0.72	0.50	No	Yes	View model
39	TTM-R3-FTHFCode IBM TSFM	Fine-tuned	37.78	40.29	0.71	0.51	No	Yes	View model
40	VISIT-2.0	Pretrained	38.56	35.82	0.71	0.50	No	No	—
41	Xihe-maxHF Ant	Zero-shot	39.91	37.29	0.71	0.49	No	No	—
42	TTM-R3-PTHFCode IBM TSFM	Pretrained	41.21	44.25	0.72	0.52	No	Yes	View model
43	Reverso-SmallHFCode MIT	Zero-shot	41.42	69.60	0.73	0.63	No	Yes	—
44	FlowState-9.1MHFCode IBM Research	Zero-shot	41.60	35.97	0.73	0.50	No	Yes	—
45	TEMPO_ENSEMBLEHF Melady Lab @ USC	Fine-tuned	42.39	34.73	0.79	0.46	Yes	No	—
46	A Kairos_50mHFCode ShanghaiTech University	Zero-shot	43.10	53.33	0.74	0.55	No	Yes	View model
47	Moirai2HFCode Salesforce AI Research	Pretrained	44.79	42.81	0.73	0.52	No	Yes	View family
48	Toto_Open_Base_1.0HFCode Datadog	Zero-shot	46.44	41.24	0.75	0.52	No	Yes	View model
49	A Kairos_23mHFCode ShanghaiTech University	Zero-shot	46.80	54.98	0.75	0.55	No	Yes	View model
50	A Kairos_10mHFCode ShanghaiTech University	Zero-shot	47.30	56.08	0.75	0.55	No	Yes	View model
51	TTM-R2-FinetunedHFCode IBM Research	Fine-tuned	47.64	57.14	0.76	0.58	Yes	Yes	View model
52	chronos_bolt_baseHFCode AWS AI Labs	Pretrained	47.85	48.98	0.81	0.57	Yes	Yes	View model
53	Toto-2.0-4mHFCode Datadog	Pretrained	47.96	43.88	0.76	0.52	No	Yes	View model
54	timesfm_2_0_500mHFCode Google Research	Pretrained	49.49	48.73	0.76	0.55	Yes	Yes	View model
55	sundial_base_128mHFCode Tsinghua University	Zero-shot	51.08	55.60	0.75	0.56	No	Yes	View model
56	xLSTM-MixerHF AIML Lab @ TU Darmstadt	Deep learning	52.87	40.62	0.78	0.51	No	No	—
57	Reverso-Nano MIT	Zero-shot	53.15	73.01	0.76	0.66	No	No	—
58	CleanTS-65MHFCode Shandong University	Zero-shot	53.38	45.63	0.80	0.54	No	Yes	—
59	chronos_bolt_smallHFCode AWS AI Labs	Pretrained	53.97	52.53	0.82	0.58	Yes	Yes	View model
60	YingLong_300mHFCode Alibaba	Zero-shot	55.31	49.11	0.80	0.55	No	Yes	View model
61	YingLong_110mHFCode Alibaba	Zero-shot	56.09	50.20	0.81	0.56	No	Yes	View model
62	TempoPFNHFCode University of Freiburg	Zero-shot	56.40	47.81	0.79	0.53	No	Yes	—
63	TabPFN-TSHFCode PriorLabs	Zero-shot	56.51	50.24	0.77	0.54	No	Yes	—
64	YingLong_50mHFCode Alibaba	Zero-shot	58.95	53.20	0.82	0.57	No	Yes	View model
65	Moirai_largeHFCode Salesforce AI Research	Zero-shot	60.65	54.41	0.87	0.60	No	Yes	View model
66	Lingjiang Alibaba Cloud	Zero-shot	60.92	56.07	1.00	0.62	No	No	—
67	Moirai_baseHFCode Salesforce AI Research	Zero-shot	61.34	54.32	0.90	0.61	No	Yes	View model
68	Chronos_largeHFCode AWS AI Labs	Pretrained	63.31	65.68	0.87	0.65	Yes	Yes	View model
69	Chronos_baseHFCode AWS AI Labs	Pretrained	63.73	66.26	0.88	0.65	Yes	Yes	—
70	PatchTST Princeton University	Deep learning	66.22	59.11	0.85	0.59	No	No	View model
71	FLAIRHFCode Mellon Inc.	Statistical	66.37	57.89	0.84	0.59	No	Yes	—
72	Chronos_smallHFCode AWS AI Labs	Pretrained	66.77	67.31	0.89	0.66	Yes	Yes	—
73	i_transformer Tsinghua University	Deep learning	67.88	61.27	0.89	0.62	No	Yes	—
74	YingLong_6mHFCode Alibaba	Zero-shot	68.30	62.04	0.88	0.61	No	Yes	View model
75	Super-LinearHFCode Ben-Gurion University of the Negev	Pretrained	68.81	79.36	0.86	0.74	Yes	Yes	—
76	TimesFMHFCode Google Research	Pretrained	70.32	65.53	1.08	0.68	Yes	Yes	View family
77	TFT Google Research	Deep learning	70.71	59.82	0.92	0.60	No	No	—
78	VisionTSHF Zhejiang University	Zero-shot	72.09	80.04	0.86	0.75	No	No	—
79	Moirai_smallHFCode Salesforce AI Research	Zero-shot	72.22	63.86	0.95	0.65	No	Yes	View model
80	N-BEATS ServiceNow	Deep learning	73.16	80.68	0.94	0.82	No	No	—
81	FFMHFCode IBM Research	Deep learning	73.55	67.90	1.02	0.70	No	No	—
82	Auto_Arima	Statistical	74.96	79.61	1.07	0.91	No	No	—
83	TTM-R2-PretrainedHFCode IBM Research	Pretrained	75.94	83.72	1.02	0.87	Yes	Yes	View model
84	TTM-R1-PretrainedHFCode IBM Research	Pretrained	77.85	83.94	1.08	0.89	Yes	Yes	View model
85	Auto_ETS	Statistical	78.36	82.91	1.21	7.49	No	No	—
86	Auto_Theta	Statistical	78.58	80.64	1.09	1.24	No	No	—
87	Seasonal_NaiveHFCode	Statistical	78.63	83.80	1.00	1.00	No	Yes	—
88	DLinear The Chinese University of Hong Kong	Deep learning	79.48	83.67	1.06	0.85	No	No	—
89	Crossformer Shanghai Jiao Tong University	Deep learning	79.87	71.95	2.57	1.64	No	No	—
90	DeepAR Amazon Research	Deep learning	80.73	74.37	1.34	0.85	No	No	—
91	TIDE Google Research	Deep learning	80.76	75.69	1.09	0.77	No	No	—
92	Lag-LlamaHF Morgan Stanley & Service Now	Pretrained	83.85	81.02	1.23	0.88	Yes	No	View model
93	NaiveHFCode	Statistical	84.22	88.56	1.27	1.59	No	Yes	—
94	VISIT-1.0	Pretrained	84.39	88.68	1.27	1.59	No	No	—

94 models · Aggregated live from the official GIFT-Eval leaderboard. Refreshes every 12 hours.Last refreshed Jun 13, 2026, 6:20 PM

Model landscape

GIFT-Eval is not only a TSFM leaderboard — it includes classical statistical baselines (ARIMA, ETS, Theta), deep-learning architectures (PatchTST, iTransformer, TFT), and agentic systems, so foundation-model results can be compared against the full prior art.

Pretrained: 26(28%)
Zero-shot: 32(34%)
Fine-tuned: 6(6%)
Agentic: 14(15%)
Deep learning: 10(11%)
Statistical: 6(6%)

Why robustness across datasets matters

A model that tops one leaderboard can collapse on a different frequency or domain. GIFT-Eval forces models to prove themselves across 23 grouped dataset slices covering different frequencies, horizons, and series shapes. If a model ranks well here, you can be more confident it will not surprise you when your data does not look like the training distribution.

Understanding Weighted Quantile Loss

WQL penalizes both overconfident and underconfident prediction intervals. A model with a low WQL produces forecast distributions that are well-calibrated — the 90th percentile prediction actually lands above the true value about 90% of the time. This matters for capacity planning, inventory, and any decision that depends on reliable uncertainty estimates rather than just the median forecast.

How to interpret it

—Lower Average Rank is better because the leaderboard aggregates placement across many slices.
—WQL matters when forecast calibration and uncertainty quality matter to the business.
—Use GIFT-Eval to sanity-check whether a model is robust beyond a single narrow domain.

Frequently asked questions

What is GIFT-Eval?: GIFT-Eval (General Time Series Forecasting Model Evaluation) is a probabilistic forecasting benchmark from Salesforce that evaluates time series foundation models across 23 diverse dataset groups spanning 7 domains and 10 frequencies. Models are ranked by Average Rank and Average Weighted Quantile Loss.
What does Average Rank mean in GIFT-Eval?: Average Rank is the mean placement of a model across every benchmark slice. A lower value means the model consistently finishes near the top across heterogeneous datasets rather than dominating only one slice.
When should I use GIFT-Eval instead of FEV Bench?: Use GIFT-Eval when you care about probabilistic forecast quality (calibrated uncertainty) and robustness across many different data domains, rather than just point-forecast accuracy on a single leaderboard.
Does GIFT-Eval test multivariate forecasting?: GIFT-Eval includes both univariate and multivariate slices. The leaderboard on TSFM.ai aggregates both variate types from the upstream grouped-by-univariate file, and you can filter per-slice rankings to inspect multivariate-only performance.
What is the 'test leak' column?: Some GIFT-Eval submissions are from models whose training data overlaps with the evaluation datasets. A 'Yes' in the test-leak column means the authors disclosed partial or full pretraining-data overlap; use the filter to compare only models with no known test-data leakage.
What model types appear on the leaderboard?: GIFT-Eval ranks pretrained foundation models, zero-shot models, fine-tuned models, agentic systems, classical deep-learning architectures (PatchTST, iTransformer, TFT, TiDE, N-BEATS), and statistical baselines (ARIMA, ETS, Theta). The page shows every type so you can benchmark a TSFM against classical prior art.
How often is the leaderboard refreshed?: TSFM.ai refetches the upstream Salesforce GIFT-Eval results every 12 hours via Next.js ISR. The 'last refreshed' timestamp at the bottom of the leaderboard reflects the most recent successful refresh.

Title	Date
GIFT-Eval: Salesforce's Comprehensive TSFM Benchmark	2026-02-15	Read →
The State of Multivariate Forecasting in 2025	2025-05-08	Read →
The Challenges of Benchmarking TSFMs	2024-11-05	Read →