Abstract
This technical report describes the Time Series Optimized Transformer forObservability (Toto), a new state of the art foundation model for time seriesforecasting developed by Datadog. In addition to advancing the state of the arton generalized time series benchmarks in domains such as electricity andweather, this model is the first general-purpose time series forecastingfoundation model to be specifically tuned for observability metrics. Toto wastrained on a dataset of one trillion time series data points, the largest amongall currently published time series foundation models. Alongside publiclyavailable time series datasets, 75% of the data used to train Toto consists offully anonymous numerical metric data points from the Datadog platform. In ourexperiments, Toto outperforms existing time series foundation models onobservability data. It does this while also excelling at general-purposeforecasting tasks, achieving state-of-the-art zero-shot performance on multipleopen benchmark datasets.