Timestral 1.0 209M
Timestral 209M is the largest release in Mistral's time series foundation-model family. It forecasts numeric sequences without task-specific training. It supports multiple target series, related input variables and a distribution of possible outcomes through quantile predictions.
Preview
Use numeric observations and related variables
The 24-layer model groups 32 time steps into each patch. Its attention operates across time and across variables. It can predict up to 1,024 future steps in one pass. For longer forecasts, the implementation appends each block's median prediction before producing the next block.Timestral 209M official model card
Download the forecasting checkpoint
The official model page links to the 209M weights. It does not list a Mistral serverless forecasting identifier. The Modified MIT license needs a separate review from the standard MIT license.Timestral 1.0 209M model page
Compare forecasts on unseen periods
Practical guidance. Start by comparing this model with the forecasting method already used for the same series. Measure error by horizon and check whether predicted intervals cover actual outcomes often enough. Adding variables helps only when those values are available at the time each forecast would have been made.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Preview |
| Access | Downloadable |
| Model identifiers | Timestral-209M |
| Release class | Public preview; downloadable checkpointTimestral 1.0 209M model page |
| Parameters | 209 million nominal parametersTimestral 209M official model card |
| History limit | 32,768 time steps, not language tokensTimestral 209M official model card |
| Single-pass horizon | 1,024 future time stepsTimestral 209M official model card |
| License | Modified MIT; company revenue condition appliesTimestral Modified MIT license |
Related models
- Timestral 1.0 123M
Mistral AI · Timestral
A 123M time-series model with 20 layers that predicts future values and quantiles from numeric observations.
Launch date not confirmedPreviewDownloadable - Timestral 1.0 28M
Mistral AI · Timestral
A 12-layer, 28M Timestral model that produces numeric forecasts with quantile estimates.
Launch date not confirmedPreviewDownloadable - Timestral 1.0 5M
Mistral AI · Timestral
The smallest Timestral release uses eight layers and 5M parameters for numeric time-series forecasting.
Launch date not confirmedPreviewDownloadable