Mistral AI has released Timestral, its first family of time series foundation models. This new set includes four zero-shot forecasting models, with parameter counts ranging from 5 million to 209 million.
Key points
- Mistral AI released Timestral, a family of four time series forecasting models.
- The largest Timestral model outperforms Google's TimesFM-3 on benchmarks.
- Timestral models can run on standard server CPUs without a GPU.
- The models offer open weights and inference code for developers.
The largest model in the Timestral family has achieved the top rank among pretrained, zero-shot models on the GIFT-Eval, TIME, and fev-bench benchmarks.
Timestral models can perform probabilistic forecasting on standard server CPUs, eliminating the need for GPUs during inference. The models offer open weights and inference code for developers. The largest 209M parameter model requires 837 MB in fp32 and can forecast multivariate targets jointly, accepting past-only, known-future, and categorical covariates.
By the numbers
- Largest Timestral model parameters
- 209M
- fev-bench SQL skill score
- 50.39
- TimesFM-3 fev-bench SQL skill score
- 48.99
The performance of the 209M Timestral model is notable, with Mistral reporting a fev-bench SQL skill score of 50.39, surpassing Google's TimesFM-3 score of 48.99. On GIFT-Eval, its MASE score of 0.658 leads t0-gamma by a narrow margin of 0.003. Timestral is a causal, patch-based transformer utilizing time attention and variate attention.
Unlike classical tools that fit one series at a time or deep models that train from scratch on each dataset, Timestral transfers patterns across domains. Each model outputs 21 quantiles per time step. The 209M model features 24 layers and an embedding dimension of 768, with each patch covering 32 time steps.
Sources and references
Original report
Mistral AI Releases Timestral: A 209M Forecasting Model Family That Outperforms Google’s TimesFM-3
MarkTechPost
Read on MarkTechPost (opens in a new tab)Linked in the original report
- Linkmistral.ai/news/timestral (opens in a new tab)
- Modelhuggingface.co/mistralai/Timestral-209M (opens in a new tab)
- Modelhuggingface.co/collections/mistralai/timestral (opens in a new tab)
- Codegithub.com/mistralai/timestral (opens in a new tab)
- Linkresearch.google/blog/timesfm-3-a-zero-shot-foundation-m… (opens in a new tab)
- Modelhuggingface.co/amazon/chronos-2 (opens in a new tab)
Ref 00O0FV7X
Cite this
APA
Interestana AI Editorial. (2026, October 19). Mistral AI Releases Timestral Forecasting Model Family. Interestana. https://interestana.com/articles/mistral-ai-releases-timestral-a-209m-forecasting-model-family-that-outperforms-googles-timesfm-3-00o0fv7x
MLA
“Mistral AI Releases Timestral Forecasting Model Family.” Interestana, 19 Oct. 2026, interestana.com/articles/mistral-ai-releases-timestral-a-209m-forecasting-model-family-that-outperforms-googles-timesfm-3-00o0fv7x.
Chicago
Interestana AI Editorial. “Mistral AI Releases Timestral Forecasting Model Family.” Interestana, October 19, 2026. https://interestana.com/articles/mistral-ai-releases-timestral-a-209m-forecasting-model-family-that-outperforms-googles-timesfm-3-00o0fv7x.
BibTeX
@misc{interestana_00o0fv7x,
author = {{Interestana AI Editorial}},
title = {{Mistral AI Releases Timestral Forecasting Model Family}},
howpublished = {Interestana},
year = {2026},
month = oct,
url = {https://interestana.com/articles/mistral-ai-releases-timestral-a-209m-forecasting-model-family-that-outperforms-googles-timesfm-3-00o0fv7x},
note = {Summary of reporting by MarkTechPost, https://www.marktechpost.com/2026/10/18/mistral-timestral-time-series-foundation-model/}
}Credit the source
Original reporting: MarkTechPost, “Mistral AI Releases Timestral: A 209M Forecasting Model Family That Outperforms Google’s TimesFM-3”, https://www.marktechpost.com/2026/10/18/mistral-timestral-time-series-foundation-model/(opens in a new tab)
Follow this story
Get Artificial Intelligence in your inbox — new stories in 40 words or fewer, once a week.
Subscribe