Simulation · Google Research
Forecast a series it has never seen, with no fitting and no training on your data.
1–30s
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
About this model
A decoder-only foundation model for time series. You give it the history and it returns the next points with a prediction interval — no seasonality to specify, no per-series model to fit, and nothing trained on what you submit. It sees a bare sequence of numbers, so anything the series cannot express is invisible to it: a policy change, a changed assay, a holiday. Read the median as what the recent past implies, and the band as the model's own uncertainty rather than a statement about the world.
Standardized I/O contract
Every model in the Hub speaks the same contract, which is what lets the Router and the agent call any of them without special-casing.
Inputs
Observed series
Steps ahead
Outputs
History, forecast and the 10th–90th percentile band on one axis.
Point forecast and the 10th, 50th and 90th percentiles per step.
Observations used, horizon and the range of the forecast.
