Timestral 1.0 5M
Timestral 5M is the smallest downloadable Timestral release. It has eight layers and a 192-dimensional internal representation. It can use a long numeric history and related variables while keeping the parameter count much lower than the other family members.
Preview
Use numeric observations and related variables
The model uses 32-step patches with missing-value indicators. It produces 21 quantiles and can forecast 1,024 steps in one pass. Longer horizons are built from successive forecast blocks. CPU and GPU inference are supported through the published PyTorch implementation.Timestral 5M official model card
Download the forecasting checkpoint
The official model page links to the 5M 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 5M model page
Compare forecasts on unseen periods
Practical guidance. Use 5M as a small baseline for an application with many short forecasts. Check both common periods and abrupt changes. A quick forecast that misses rare demand peaks can be less useful than a slower one, so compare the decision the forecast supports as well as average error.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Preview |
| Access | Downloadable |
| Model identifiers | Timestral-5M |
| Release class | Public preview; downloadable checkpointTimestral 1.0 5M model page |
| Parameters | 5 million nominal parametersTimestral 5M official model card |
| History limit | 32,768 time steps, not language tokensTimestral 5M official model card |
| Single-pass horizon | 1,024 future time stepsTimestral 5M 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 209M
Mistral AI · Timestral
The 209M Timestral model forecasts numeric series with a 24-layer architecture and quantile outputs.
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