Skip to content
Mistral· LaunchAvailable

Mistral Releases Timestral: Open Models for Time Series Forecasting

Mistral AI is expanding its portfolio with the launch of Timestral, a set of four compact time series forecasting models with up to 209 million parameters that can run directly on standard CPUs.

By the Aheadline editorial team·20 okt. 2026·1 min read·Source: Entity-watch: Mistral AIVerifierad signalAI-generated
Mistral Releases Timestral: Open Models for Time Series Forecasting
Mistral Releases Timestral: Open Models for Time Series Forecasting
Mistral Releases Timestral: Open Models for Time Series Forecasting
By · Verktygs- & infrastrukturreporter
Last updated
Vad betyder det för mig?

What happened?

Mistral AI has launched Timestral, the company's first family of foundational models for time series forecasting. The family consists of four sizes: 5 million, 28 million, 123 million, and 209 million parameters. They handle multivariate time series with probabilistic zero-shot forecasting, a context length of 32,768 steps, and 21 output quantiles.

Key facts

Lanseringsdatum15 oktober 2026
Modellstorlekar5M, 28M, 123M och 209M parametrar
Kontextfönster32 768 steg
Utdatakvantiler21 stycken
LicensformModifierad MIT-licens

Why it matters

According to Mistral, the top-tier 209-million-parameter model performs best among pretrained zero-shot models across five metrics on the GIFT-Eval, TIME, and fev-bench tests, despite having roughly one-third fewer parameters than previous leading models. The 28-million-parameter model is also reported to outperform the competitor Chronos-2 on GIFT-Eval and TIME, despite being approximately one-quarter the size.

Who is affected?

The launch is relevant to developers, data scientists, and technical teams working with time series data and machine learning. The compact sizes make the models particularly well-suited for organisations that need to perform computations locally on standard CPUs.

Impact on the EU

As a European model family developed by the French firm Mistral AI, Timestral is released under a modified MIT licence and is immediately available to developers within the EU.

What else you should know

The models are designed to run on CPUs. On a server with eight threads, a one-day forecast takes approximately 10 milliseconds with the 5M model and about 140 milliseconds with the 209M model.

Frequently asked questions

Quick answers about this story

Vad har hänt?
Mistral AI har släppt Timestral, en familj av fyra öppna grundmodeller för tidsserieprognoser med upp till 209 miljoner parametrar.
När hände det?
Timestral lanserades den 15 oktober 2026.
Varför spelar det roll?
Modellerna levererar ledande zero-shot-prestanda på etablerade benchmarks trots ett avsevärt lägre parameterantal och kan köras snabbt direkt på vanliga processorer.
Vilken licens har Timestral?
Modellerna distribueras under en modifierad MIT-licens.
Original source
Entity-watch: Mistral AI·techtools.cz

The link opens in a new window and leads to the publisher's own site.

Verifierad signal

Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.

AI-verktyg i artikeln

Topics

#Mistral AI#Models
[ STAY UP TO DATE ]

Get similar news straight to your inbox

No affiliate linksCancel anytimeGDPR-friendly
[ Frequency ]
[ What do you want to read about? ]

You'll receive updates on 2 topics.

The reader's room

Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.

Sign in to submit a comment or question.

Loading comments…
How this affects you

Read the article through your role

  • Decide whether this affects strategy over 6–12 months or is just noise.
  • Discuss with leadership: do we own the right question or does ownership need to move?
  • Ask: what risk are we taking by NOT acting on this this quarter?

Generated angle — not editorial analysis of "Mistral Releases Timestral: Open Models for Time Series Fore"