Topic
Transformer-based neural networks that forecast time-series data, including pretrained zero-shot foundation models such as Google's TimesFM.
Mistral released Timestral, four CPU-capable forecasting models, with weights licensed only to firms at or under $20M in monthly revenue. The license text, as DEV Community quotes it, tests the preceding month, so eligibility can change monthly.
Reality
Databricks added four SQL-native predictive functions in Beta and upgraded ai_forecast, according to its own blog post. Analysts can run them without training a model, but the post puts no number on the accuracy gain.
Reality
Ant International says its upgraded forecasting model is in production at Citi, HSBC, Deutsche Bank, Standard Chartered and Barclays, with a claim of 60% lower hedging costs attached.
Perspective Coverage
3 publishersReality
BMW Group's CLEA forecasts hundreds of thousands of account-service pairs every day with Prophet on Lambda, then has to decide which of those deviations is worth an account owner's email. Compute runs about $50 a month.
Reality
Google's new forecaster is 330m parameters trained on a trillion time points. In zero-shot multivariate mode it anticipates a 20% promotion lift that a univariate model misses. That is most of what an in-house demand model does.
Reality
Google's new 330 million parameter forecaster takes covariates and related series zero-shot in one forward pass, which puts a lot of bespoke per-SKU plumbing up for review, assuming the accuracy transfers to your own series.
Reality