Google AI Releases TimesFM 3 Forecasting Model

Google AI has released TimesFM 3, a 330 million-parameter foundation model designed for zero-shot forecasting of multivariate time series.

A model for multivariate forecasting

TimesFM 3 is described as a foundation model for time-series forecasting, with the ability to work across multiple variables and produce forecasts without task-specific training. Its zero-shot design is the central feature identified in the available announcement.
Multivariate time-series forecasting involves analyzing several related measurements over time. However, the available source material does not provide further details about the model’s architecture, training data, benchmarks, supported input formats, or access requirements.

What to watch next

The release positions TimesFM 3 within Google AI’s work on general-purpose forecasting models. The model’s 330 million parameters indicate a substantial expansion in scale, but no performance comparisons or deployment examples are provided.
Further information would be needed to assess how TimesFM 3 performs against other forecasting systems, which industries it targets, and whether it is available for public or commercial use.