{"name":"timestral","repo":"mistralai/timestral","url":"https://midorreal.com/project/mistralai-timestral","repository":"https://github.com/mistralai/timestral","description":"Repository for Timestral, the time-series foundation model by Mistral AI","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/mistralai/timestral","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/mistralai/timestral","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/mistralai/timestral/releases","checked":"21 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/mistralai/timestral/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/mistralai/timestral","checked":"22 Nov 2026"}},"stars":1244.0,"writeup":{"what_it_is":"Timestral is a family of time series foundation models for multivariate probabilistic forecasting. This repository provides standalone inference code for Timestral models, which needs PyTorch, NumPy, safetensors and huggingface_hub and runs on CPU or GPU.","audience":"Developers and researchers who need time series forecasting with pretrained models.","claims":[{"kind":"specific","claim":"Timestral-209M ranked first among pretrained and zero-shot models on GIFT-Eval, TIME and fev-bench at launch.","excerpt":"At launch, Timestral-209M ranked first among pretrained and zero-shot models on the three main public forecasting benchmarks","status":"not_checked"},{"kind":"specific","claim":"Forecasts produce 21 quantiles per time step.","excerpt":"**Probabilistic**: 21 quantiles per time step.","status":"not_checked"},{"kind":"specific","claim":"Context can be up to 32,768 steps long.","excerpt":"**Long context**: up to 32,768 steps.","status":"not_checked"},{"kind":"specific","claim":"Inference runs on CPU without a GPU.","excerpt":"**CPU-friendly**: no GPU needed for inference.","status":"not_checked"},{"kind":"specific","claim":"Variates of a series are forecast jointly and support past-only and known-future covariates.","excerpt":"- **Multivariate**: the variates of a series are forecast jointly.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-01"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":65,"reality":1,"gap":64,"momentum":0.2,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":{"estimate":true,"lean":"too_close","gap_low":22,"gap_high":58,"rule":"early-2.0"},"changes":[{"date":"2026-10-15T10:20:18+00:00","text":"Repository created","source":"https://github.com/mistralai/timestral"}],"cite":"timestral is a GitHub project at https://github.com/mistralai/timestral written mainly in Python, described there as \"Repository for Timestral, the time-series foundation model by Mistral AI\".","attribution":"Data from Mid or Real, https://midorreal.com/project/mistralai-timestral, checked 22 November 2026. Attribution with a link is required.","still_real":null}