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·112 models · 112 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.

CastStar ranks first with an average MASE rank of 16.44 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 112 of 112 models · 38 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	CastStarHF USTC-AGI	Agentic	16.44	15.02	0.66	0.45	No	No	—
2	Falcon-Agent ant-intl	Agentic	16.67	17.27	0.67	0.45	No	No	—
3	RacineCast-1-NCHFCode Racine.ai	Agentic	18.68	18.80	0.67	0.46	No	Yes	—
4	RacineCast-1HFCode Racine.ai	Agentic	18.72	19.53	0.67	0.46	No	Yes	—
5	Cobra-AgentHF Dalpha AI	Agentic	20.80	20.27	0.68	0.46	No	No	—
6	Toto-2.0-FnFHFCode Datadog	Agentic	22.25	22.15	0.68	0.46	No	Yes	—
7	PrismHF Birla AI Labs	Agentic	22.79	22.51	0.68	0.47	No	No	—
8	RAES-Conductance-EnsembleHFCode California State University Northridge	Agentic	22.97	23.32	0.66	0.46	Yes	Yes	—
9	HistRoute-CV Hitsz iLearn-ML	Agentic	23.79	24.18	0.68	0.47	No	No	—
10	Taichu-TimeSeries-AgentHFCode zidongtaichu	Agentic	24.03	25.70	0.67	0.46	No	No	—
11	metis-autocastHF Metis	Agentic	24.51	21.26	0.68	0.46	No	No	—
12	Toto-2.0-2.5B-FTHFCode Datadog	Fine-tuned	25.26	24.26	0.68	0.46	No	Yes	—
13	TSOrchestraHFCode Melady Lab @ USC	Agentic	25.33	23.94	0.68	0.47	No	No	—
14	DeOSAlphaTimeGPTPredictor-2025HF vencortex®	Zero-shot	26.19	24.75	0.68	0.47	No	No	—
15	TimeRouterHFCode UConn & Morgan Stanley	Agentic	27.48	30.92	0.67	0.47	No	Yes	—
16	RAES-Conductance-Ensemble-VHFCode SETI	Agentic	27.54	27.93	0.67	0.46	No	Yes	—
17	MoiraiAgent-leakingHFCode Salesforce AI Research	Agentic	28.13	31.64	0.68	0.47	Yes	Yes	—
18	MoiraiAgentHFCode Salesforce AI Research	Agentic	30.32	32.96	0.69	0.48	No	Yes	—
19	CredenceHFCode ContinualIST	Agentic	30.43	27.80	0.69	0.47	No	Yes	—
20	Falcon-2.0HFCode ant-intl	Pretrained	31.60	41.21	0.67	0.49	No	Yes	—
21	STRIDE (+Chronos-2)HF Google Cloud AI Research	Pretrained	33.10	27.91	0.67	0.45	No	No	—
22	Samay Kairosity	Agentic	33.29	30.36	0.70	0.48	No	No	—
23	TiRex-2-PretrainedHFCode NXAI	Pretrained	33.66	32.19	0.68	0.47	No	Yes	—
24	Toto-2.0-2.5BHFCode Datadog	Pretrained	34.31	34.13	0.70	0.48	No	Yes	View model
25	ZooCast-Top1Code	Agentic	34.42	35.18	0.69	0.48	No	No	—
26	Migas-1.0HF Synthefy	Pretrained	34.46	81.23	0.70	0.60	No	No	—
27	STRIDE (+Timer-S1)HF Google Cloud AI Research	Pretrained	34.88	32.86	0.67	0.46	No	No	—
28	TimeCopilotHFCode	Agentic	35.29	33.34	0.71	0.48	No	Yes	—
29	SynapseHF Google Cloud AI Research	Agentic	35.58	31.78	0.70	0.48	No	No	—
30	Toto-2.0-1BHFCode Datadog	Pretrained	36.19	35.32	0.70	0.48	No	Yes	View model
31	Toto-2.0-313mHFCode Datadog	Pretrained	37.36	35.88	0.70	0.48	No	Yes	View model
32	Chronos-2HFCode AWS	Pretrained	37.59	38.04	0.70	0.49	No	Yes	View model
33	TurkForecast-FM-Chronos2-LoRA-v1HFCode TurkForecast (Mert Karatay)	Fine-tuned	38.68	37.98	0.66	0.42	No	Yes	—
34	EFG-base LAIR	Zero-shot	39.49	38.73	0.70	0.48	No	No	—
35	TSOrchestra-testHF Melady Lab @ USC	Fine-tuned	39.74	30.03	0.70	0.47	Yes	No	—
36	TiRex-2-ZeroshotHFCode NXAI	Zero-shot	40.40	35.60	0.70	0.48	No	Yes	—
37	ZeusHFCode GestaltCog Lab @ ICT	Pretrained	40.81	36.01	0.69	0.48	No	Yes	—
38	PatchTST-FM-r1HFCode IBM TSFM & Rensselaer Polytechnic Institute	Zero-shot	41.26	37.90	0.71	0.48	No	Yes	View model
39	Granite-FlowState-r1.1HFCode IBM TSFM	Zero-shot	41.55	42.43	0.70	0.49	No	Yes	—
40	Timer-S1HFCode Tsinghua & ByteDance	Pretrained	41.61	42.35	0.69	0.49	No	Yes	View model
41	ValBestSingle-cmttHFCode N/A	Agentic	41.78	82.64	0.71	0.60	No	No	—
42	TimesFM-2.5HFCode Google Research	Zero-shot	42.11	41.35	0.71	0.49	No	Yes	View model
43	FlowState-r1.1HFCode IBM TSFM	Zero-shot	42.43	41.10	0.70	0.49	No	Yes	View model
44	LongSeer-v1.0 LongShine AI Research	Zero-shot	44.45	44.72	0.71	0.49	No	No	—
45	Falcon-XHFCode ant-intl	Pretrained	45.40	44.94	0.69	0.49	No	Yes	—
46	TiRexHFCode NXAI	Zero-shot	46.18	39.54	0.72	0.49	No	Yes	View model
47	Granite-PatchTST-FM-r1HFCode IBM TSFM & Rensselaer Polytechnic Institute	Zero-shot	47.00	41.41	0.72	0.49	No	Yes	View model
48	Reverso MIT	Zero-shot	47.16	83.61	0.71	0.61	No	No	—
49	Toto-2.0-22mHFCode Datadog	Pretrained	47.76	42.99	0.72	0.50	No	Yes	View model
50	Xihe-ultraHF Ant	Zero-shot	48.53	46.33	0.70	0.49	No	No	—
51	Chronos-2-SynthHFCode AWS	Zero-shot	48.61	47.29	0.72	0.50	No	Yes	View model
52	TTM-R3-FTHFCode IBM TSFM	Fine-tuned	49.44	52.60	0.71	0.51	No	Yes	View model
53	VISIT-2.0	Pretrained	49.91	46.96	0.71	0.50	No	No	—
54	t0-alphaHFCode The Forecasting Company	Pretrained	50.95	42.06	0.72	0.49	No	Yes	—
55	TEMPO_ENSEMBLEHF Melady Lab @ USC	Fine-tuned	51.66	43.46	0.79	0.46	Yes	No	—
56	Xihe-maxHF Ant	Zero-shot	51.71	48.65	0.71	0.49	No	No	—
57	TTM-R3-PTHFCode IBM TSFM	Pretrained	53.35	57.21	0.72	0.52	No	Yes	View model
58	FlowState-9.1MHFCode IBM Research	Zero-shot	53.63	47.64	0.73	0.50	No	Yes	—
59	Reverso-SmallHFCode MIT	Zero-shot	53.86	86.30	0.73	0.63	No	Yes	—
60	A Kairos_50mHFCode ShanghaiTech University	Zero-shot	55.56	67.63	0.74	0.55	No	Yes	View model
61	Moirai2HFCode Salesforce AI Research	Pretrained	57.31	54.95	0.73	0.52	No	Yes	View family
62	Toto_Open_Base_1.0HFCode Datadog	Zero-shot	59.58	53.47	0.75	0.52	No	Yes	View model
63	A Kairos_23mHFCode ShanghaiTech University	Zero-shot	59.99	69.52	0.75	0.55	No	Yes	View model
64	chronos_bolt_baseHFCode AWS AI Labs	Pretrained	60.54	61.93	0.81	0.57	Yes	Yes	View model
65	A Kairos_10mHFCode ShanghaiTech University	Zero-shot	60.71	70.60	0.75	0.55	No	Yes	View model
66	TTM-R2-FinetunedHFCode IBM Research	Fine-tuned	60.76	71.98	0.76	0.58	Yes	Yes	View model
67	CHARMHFCode C3 AI	Zero-shot	61.21	27.77	0.76	0.48	No	No	—
68	Toto-2.0-4mHFCode Datadog	Pretrained	61.40	56.73	0.76	0.52	No	Yes	View model
69	timesfm_2_0_500mHFCode Google Research	Pretrained	62.08	61.56	0.76	0.55	Yes	Yes	View model
70	sundial_base_128mHFCode Tsinghua University	Zero-shot	64.46	69.94	0.75	0.56	No	Yes	View model
71	xLSTM-MixerHF AIML Lab @ TU Darmstadt	Deep learning	66.41	52.34	0.78	0.51	No	No	—
72	CleanTS-65MHFCode Shandong University	Zero-shot	67.31	58.89	0.80	0.54	No	Yes	—
73	Reverso-Nano MIT	Zero-shot	67.40	90.15	0.76	0.66	No	No	—
74	chronos_bolt_smallHFCode AWS AI Labs	Pretrained	67.66	66.33	0.82	0.58	Yes	Yes	View model
75	PatchFMHFCode LITIS	Zero-shot	69.54	63.24	0.81	0.56	No	Yes	—
76	YingLong_300mHFCode Alibaba	Zero-shot	70.00	63.11	0.80	0.55	No	Yes	View model
77	YingLong_110mHFCode Alibaba	Zero-shot	70.74	64.15	0.81	0.56	No	Yes	View model
78	TabPFN-TSHFCode PriorLabs	Zero-shot	70.86	63.84	0.77	0.54	No	Yes	—
79	TempoPFNHFCode University of Freiburg	Zero-shot	71.00	61.10	0.79	0.53	No	Yes	—
80	YingLong_50mHFCode Alibaba	Zero-shot	74.00	67.57	0.82	0.57	No	Yes	View model
81	Lingjiang Alibaba Cloud	Zero-shot	74.70	70.07	1.00	0.62	No	No	—
82	Moirai_largeHFCode Salesforce AI Research	Zero-shot	75.30	68.02	0.87	0.60	No	Yes	View model
83	Moirai_baseHFCode Salesforce AI Research	Zero-shot	76.16	68.06	0.90	0.61	No	Yes	View model
84	Metamorph1.0 SRI International	Pretrained	76.25	65.61	0.82	0.55	No	No	—
85	LiteSpecFormerHFCode FlowVortex (Xidian University)	Zero-shot	76.65	72.89	0.87	0.62	No	Yes	—
86	Chronos_largeHFCode AWS AI Labs	Pretrained	78.65	81.29	0.87	0.65	Yes	Yes	View model
87	Chronos_baseHFCode AWS AI Labs	Pretrained	79.31	82.00	0.88	0.65	Yes	Yes	—
88	PatchTST Princeton University	Deep learning	81.90	74.20	0.85	0.59	No	No	View model
89	FLAIRHFCode Mellon Inc.	Statistical	82.05	72.40	0.84	0.59	No	Yes	—
90	Chronos_smallHFCode AWS AI Labs	Pretrained	82.44	83.00	0.89	0.66	Yes	Yes	—
91	i_transformer Tsinghua University	Deep learning	83.87	76.65	0.89	0.62	No	Yes	—
92	YingLong_6mHFCode Alibaba	Zero-shot	84.69	77.69	0.88	0.61	No	Yes	View model
93	Super-LinearHFCode Ben-Gurion University of the Negev	Pretrained	84.96	96.87	0.86	0.74	Yes	Yes	—
94	TimesFMHFCode Google Research	Pretrained	86.64	81.24	1.08	0.68	Yes	Yes	View family
95	TFT Google Research	Deep learning	86.92	74.79	0.92	0.60	No	No	—
96	VisionTSHF Zhejiang University	Zero-shot	88.77	97.61	0.86	0.75	No	No	—
97	Moirai_smallHFCode Salesforce AI Research	Zero-shot	88.84	79.44	0.95	0.65	No	Yes	View model
98	N-BEATS ServiceNow	Deep learning	89.67	98.26	0.94	0.82	No	No	—
99	FFMHFCode IBM Research	Deep learning	90.09	83.74	1.02	0.70	No	No	—
100	Auto_Arima	Statistical	91.51	96.77	1.07	0.91	No	No	—
101	TTM-R2-PretrainedHFCode IBM Research	Pretrained	92.85	101.5	1.02	0.87	Yes	Yes	View model
102	Auto_ETS	Statistical	94.80	100.0	1.21	7.49	No	No	—
103	TTM-R1-PretrainedHFCode IBM Research	Pretrained	94.88	101.8	1.08	0.89	Yes	Yes	View model
104	Auto_Theta	Statistical	95.09	97.45	1.09	1.24	No	No	—
105	Seasonal_NaiveHFCode	Statistical	95.61	101.3	1.00	1.00	No	Yes	—
106	Crossformer Shanghai Jiao Tong University	Deep learning	95.81	86.77	2.57	1.64	No	No	—
107	DLinear The Chinese University of Hong Kong	Deep learning	96.70	101.4	1.06	0.85	No	No	—
108	DeepAR Amazon Research	Deep learning	97.92	90.92	1.34	0.85	No	No	—
109	TIDE Google Research	Deep learning	97.94	92.65	1.09	0.77	No	No	—
110	Lag-LlamaHF Morgan Stanley & Service Now	Pretrained	101.5	98.51	1.23	0.88	Yes	No	View model
111	NaiveHFCode	Statistical	101.7	106.3	1.27	1.59	No	Yes	—
112	VISIT-1.0	Pretrained	101.8	106.5	1.27	1.59	No	No	—

112 models · Aggregated live from the official GIFT-Eval leaderboard. Refreshes every 12 hours.Last refreshed Jul 29, 2026, 6:05 AM

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: 31(28%)
Zero-shot: 37(33%)
Fine-tuned: 6(5%)
Agentic: 22(20%)
Deep learning: 10(9%)
Statistical: 6(5%)

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 →