AINew modelStory 03
Open models forecast long time series
What changedTimestral is a family of four open forecasting models, from 5 million to 209 million parameters. Its design separates future prediction from the observed context and accepts known future signals such as holidays. The authors report that the two largest models lead three forecasting benchmarks.
The useful part
Why it matters
Teams can try one model family across demand, finance or operations data without building a bespoke forecaster for every horizon.
Keep in mind
Good to know
The results are author-reported benchmarks; performance on noisy production series and the true operating cost still need independent testing.
Evidence