Re:Timestral: Mistral release time-series foundation models
by GeoffNN on Monday October 19, 2026 @07:54AM
(#50151269)
Cofounder of The Forecasting Company here - Check ours! It's open-weights: https://huggingface.co/theforecastingcompany/t0-gamma
[ Parent
| on HN ]
Re:Timestral: Mistral release time-series foundation models
by dfdydx on Monday October 19, 2026 @08:02AM
(#50151344)
So is Mistral's model, right? Genuine question because I had been following your company - is there any use case where t0 is better than timestral?
Re:Timestral: Mistral release time-series foundation models
by GeoffNN on Monday October 19, 2026 @08:43AM
(#50151727)
Ah thanks!
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. Ours is Apache 2.
If you compare the 2, it's not a strict "timestral is better" story - the winrates are tight - your probably should try both on your use-case. We have a 42% winrate vs Timestral on fevbench for instance.
Ofc, our main value is not just these models, but also 1) extra capabilities (like our model that can natively account for stockouts/censored data, and others coming soon), 2) our systems and products around the model, and 3) we're all laser focused on time-series/forecasting
Re:Timestral: Mistral release time-series foundation models
by gturri on Monday October 19, 2026 @09:52AM
(#50152525)
interesting.
The github repo (https://github.com/mistralai/timestral ) says the license is Apache 2. Where does the 20M$ revenue limit comes from?
Re:Timestral: Mistral release time-series foundation models
by orra on Monday October 19, 2026 @10:27AM
(#50152939)
It's in the licence of the weights, which you can see on Huggingface.