The announcement introduces Toto, a state-of-the-art time series foundation model, and BOOM, a new public observability benchmark, both of which are open source and available under the Apache 2.0 license. Toto, trained solely on Datadog's internal telemetry metrics, excels in performance on the BOOM benchmark, and other established benchmarks like GIFT-Eval and LSF, by effectively handling the unique challenges posed by observability metrics such as sparsity, spikes, and high-cardinality multivariate series. The BOOM benchmark, which comprises 350 million observations across 2,807 real-world time series, is designed to evaluate the performance of models on observability metrics, which are crucial for operations like anomaly detection and predictive forecasting. Toto's architecture incorporates innovative features such as a Student-t mixture model prediction head and a patch-based causal normalization approach, allowing it to generalize well to nonstationary data and achieve superior performance. The availability of these tools aims to advance the field of time series analysis and encourage community engagement, with Datadog seeking to expand its AI research team.