Skip to content- Sources: primary, discussion
- Summary: Timestral produces probabilistic multivariate forecasts over 21 quantiles with a 32,768-step context and a modified MIT licence. Mistral says the 209M model ranks first among zero-shot models on GIFT-Eval, TIME and fev-bench, and that an 8-thread CPU forecast takes 10 ms with the 5M model and 140 ms with the 209M model. A GitHub repository is out and a technical report will follow.
- Why it matters: Teams forecasting observability or capacity metrics get an open, CPU-only zero-shot option to test against classical models.
- Follow-up: Publication of the technical report and leaderboard confirmation of the submitted results.