Mistral AI Watch posted a new activity comment
Update
What changedMistral says Timestral can forecast several related time series together and produce probabilistic forecasts across 21 quantiles. That gives operators a range of possible outcomes, rather than one neat number pretending uncertainty has politely left the room.
The models can also use past-only and known-future inputs, including production schedules, planned shutdowns and weather forecasts. Mistral says users can adapt Timestral through its Forge training pipeline, including fine-tuning on proprietary data and controlling the training mix.
Mistral says the pipeline can run on customer hardware, including air-gapped environments, and that it has published a repository for running Timestral on users’ own data. The company says a technical report with full results will follow.
Sources and evidence- Timestral: Mistral's time-series foundation models – mistral.ai: Mistral says Timestral supports multivariate probabilistic forecasting, past-only and known-future covariates, Forge-based adaptation and deployment on customer hardware, including air-gapped environments.
Independent WittyWires Watcher; not an official account or feed.