Timestral by Mistral
Time-series foundation model in AI tools
Free open-weight time-series forecasting models that people find strong on benchmarks and can be fine-tuned on their own data. Main caveat: a restrictive license requires a commercial agreement above $20M monthly revenue.
Free open-weight time-series forecasting models that people find strong on benchmarks and can be fine-tuned on their own data. Main caveat: a restrictive license requires a commercial agreement above $20M monthly revenue.
21 people gave an opinion
- recommend it
- 10
- criticize it
- 11
- approval share
- 48%
Recommended in 1 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: not recommended
- May 2026: not recommended
- June 2026: not recommended
- July 2026: not recommended
- August 2026: not recommended
- September 2026: not recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 10 people recommend itNeeds 10 or more, met.
- 0.9 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 1 of 13 monthsNeeds 4 or more, not met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Free, open-weight models
“they let you build a $200M yearly revenue company on top of it for free, seems quite generous”
Comment from 20 Oct 2026 in “Timestral: Mistral's time-series foundation models” “It runs GPUless and is open source, so you're just paying the normal compute cost.”
Comment from 21 Oct 2026 in “Timestral: Mistral's time-series foundation models” “Timestral is free and open weights.”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” Strong forecasting and benchmark results
“are best-in-class for each model size, and are best-in-class overall based on the public benchmarks”
Comment from 16 Oct 2026 in “Timestral” “this is a great time series forecasting model, we are migrating to this in our pod”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “get apparently quite good results (and fine-tune on my data, it seems)”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” Can fine-tune on own data
“this might be very good for feature selection to refine a much more specialized model”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “get apparently quite good results (and fine-tune on my data, it seems)”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models”
What people warn about
Each mark is one comment
Restrictive license above revenue threshold
“unfortunate about the restrictive license”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “The Mistral model has a restrictive license, where you need a specific commercial agreement to use it if you make more than 20M$ revenue/month.”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “it has a "Modified" MIT, which requires you to buy from Mistral if your global business makes more than $20m a month.”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” - Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models”
- Comment from 16 Oct 2026 in “Timestral open weight timeseries foundation models from Mistral”
- Comment from 21 Oct 2026 in “Timestral: Mistral's time-series foundation models”
Fails to beat simple baselines
“you're going to get completely beaten by a comparatively simple system that knows the time of day”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “It did not beat our AutoARIMA baselines, they are pretty noisy series though.”
Comment from 20 Oct 2026 in “Timestral: Mistral's time-series foundation models” “These are essentially numerology models.”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” Unclear real-world use cases
“I like that it's something different but I'm struggling to think of a way to use this.”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models” “Real-world usability is very limited, which is probably the reason why the blog post features an imaginary customer”
Comment from 19 Oct 2026 in “Timestral: Mistral's time-series foundation models”
What people said
All 22 comments that mention it, newest first. Each links to the original comment.
“It runs GPUless and is open source, so you're just paying the normal compute cost.”
“ignore their stupid not open license”
“It did not beat our AutoARIMA baselines, they are pretty noisy series though.”
“It does make it difficult to deploy if you are part of an enterprise”
“it says '$', but doesn't even say WHICH '$'.”
“they let you build a $200M yearly revenue company on top of it for free, seems quite generous”
“You are not authorized to exercise any rights under this license if the global consolidated monthly revenue”
“I like that it's something different but I'm struggling to think of a way to use this.”
“this is a great time series forecasting model, we are migrating to this in our pod”
“unfortunate about the restrictive license”
“this is some great european sovereignty right there!”
“it has a "Modified" MIT, which requires you to buy from Mistral if your global business makes more than $20m a month.”
“Timestral is free and open weights.”
“The Mistral model has a restrictive license, where you need a specific commercial agreement to use it if you make more than 20M$ revenue/month.”
“you're going to get completely beaten by a comparatively simple system that knows the time of day”
“this might be very good for feature selection to refine a much more specialized model”
“These are essentially numerology models.”
“Real-world usability is very limited, which is probably the reason why the blog post features an imaginary customer”
“crazy how they make an improvement of 0.03 look like a 5x difference”
“get apparently quite good results (and fine-tune on my data, it seems)”
“basically the same $20M permissioned license clause Meta is putting on their models”
“are best-in-class for each model size, and are best-in-class overall based on the public benchmarks”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.