Picture published by the creator on Hugging Face.
Timestral-209M: Forecast future values from numeric history
Added · Updated
Made by Mistral
View profile on Hugging Face (opens in a new tab)Forecast future values from numeric time-series data, including covariates and uncertainty ranges.
For analysts using Python on a CPU or GPU.
Model details
Original model mistralai/Timestral-209M
Input Unknown Output Unknown
Ways to use it Run locally
What it does
Timestral-209M predicts numeric time series without fine-tuning for a particular dataset. It accepts one or more target series, optional past-only or known-future covariates, and can mark missing observations. The model returns 21 quantile forecasts, which show a range of possible future values instead of only one estimate. The 209-million-parameter model uses a patch-based Transformer and can take up to 32,768 historical time steps, then forecast up to 1,024 steps in one pass. The card says inference works on CPU or GPU.
How to use it
Install the Timestral Python package and PyTorch, load mistralai/Timestral-209M with TimestralForecaster, then pass a numeric series and the requested prediction length to forecast(). Read the median forecast or select quantiles for lower and upper estimates.
Common use cases
- Enter a daily sales series and request a future horizon; receive median and interval forecasts to inspect.
- Provide several related measurements together and forecast their future values.
- Include known future calendar or planned values as covariates and compare the returned forecast ranges.
Can I use this?
Requirements Python 3.10 or later, PyTorch 2.4 or later, and numeric time-series input; the card gives a CPU-or-CUDA example. The model is under a Modified MIT License that includes an attribution requirement and a stated company-revenue eligibility threshold.
Caveat The card’s benchmarks are Mistral’s reported results and do not guarantee fit for a reader’s data. The Modified MIT License includes attribution and company-revenue conditions; consult its linked terms.
Worth knowing The model card names a Modified MIT License. The license requires attribution and limits use by companies above its stated monthly-revenue threshold; review the linked license for the exact conditions. Forecasts are model outputs, not guaranteed outcomes.
Why it's here
Not trending this week
Last seen on the Hugging Face trending list on 27 October 2026. It stays on BuildTube so you can still find it.
- 486 people have liked it on Hugging Face.
- It was downloaded 24,440 times in the last 30 days.
Numbers checked on 8 November 2026; not refreshed since.
Behind it
View model pageBuildTube has not run or verified this project. Everything above is written from what the creator published.
Made this?
Quick questions
- What is Timestral-209M?
- Timestral-209M predicts numeric time series without fine-tuning for a particular dataset.
- What can I use Timestral-209M for?
- Enter a daily sales series and request a future horizon; receive median and interval forecasts to inspect; Provide several related measurements together and forecast their future values; Include known future calendar or planned values as covariates and compare the returned forecast ranges
- Can I run Timestral-209M on my computer?
- It runs on your own hardware with a model runtime. Python 3.10 or later, PyTorch 2.4 or later, and numeric time-series input; the card gives a CPU-or-CUDA example. Python 3.10+, PyTorch 2.4+, and numeric time-series data; CPU or GPU inference.
- What do I need to use Timestral-209M?
- Python 3.10 or later and PyTorch 2.4 or later; the model card shows CPU or CUDA inference and names Modified MIT license conditions.
- What license does Timestral-209M use?
- The model card names a Modified MIT License. The license requires attribution and limits use by companies above its stated monthly-revenue threshold; review the linked license for the exact conditions.
More from this week
- Idol Turn one picture of an object into a 3D Gaussian splat
Idol turns one photo, render or generated image of an object into a 3D Gaussian splat file.
- P-H3-Image-Comfy Make and edit 8-megapixel images in ComfyUI
Turn short text requests into 8-megapixel images and edit pictures on your own graphics card.
- agent-desktop Let AI agents read and click desktop apps
For developers building computer-use agents: agent-desktop lets an AI agent read and operate desktop apps on Mac, Windows or Linux from a command line.
Explore by topic
Post this
Download share imageFor makers
Made this? Show it’s on BuildTube — add the badge to your README or embed the card on your site.
README badge (Markdown)
[](https://buildtube.app/p/hf-mistralai-timestral-209m/)Embed card (HTML)
<iframe src="https://buildtube.app/embed/hf-mistralai-timestral-209m/" width="560" height="240" style="border:0;max-width:100%" loading="lazy" title="Timestral-209M on BuildTube"></iframe>