The development of foundation models for time series data is crucial for advancing the field of generative AI applications that cover various use cases such as chatbots, code completion, autonomous agents, image generation and more. Current large language models (LLMs) are not optimal for understanding observability metrics, which require processing numerical time series data with high frequency and precision. Dedicated foundation models like Time Series Optimized Transformer for Observability (Toto) have the potential to complement and augment the reasoning capabilities of general-purpose LLMs. Toto achieves top performance on several open time series benchmarks, consistently outperforming existing models in key accuracy metrics, making it a state-of-the-art foundation model for time series forecasting. Its training dataset is nearly a trillion data points, compiled from publicly available time series datasets, which helps make the model generalizable to other time series domains. Toto's performance surpasses that of other recent zero-shot and full-shot models in both open benchmarks and Datadog's observability benchmark, showcasing its potential for reliable and precise forecasting capabilities.