Mistral releases Timestral, forecasting models it says rank first on three public benchmarks and run on an ordinary server CPU

Mistral released Timestral on 15 October, its first family of time-series foundation models, in four sizes from 5 million to 209 million parameters. Mistral says the largest ranks first among pre-trained, zero-shot models on the GIFT-Eval, TIME and fev-bench forecasting benchmarks, and that it forecasts without a GPU: one day ahead from three weeks of 15-minute history takes 10 ms with the 5M model and 140 ms with the 209M model on a standard server CPU with 8 threads. Organisations can fine-tune it on their own data through Mistral Forge.

• Ask your developer or accountant whether you already forecast anything that repeats, such as weekly orders, stock or cash in, and whether a ready-made forecasting model could replace a spreadsheet guess.
• Treat the benchmark claims as Mistral's own and ask for a test on a year of your own figures before any decision rests on it.

Source: Mistral AI. This is our short summary of their reporting. The paragraphs marked "Our read" and the things to do are SeedFoundry's opinion. The summary is drafted with AI tools. Image: Mistral AI, from its own announcement. How we source and credit