Toto 2.0: Time Series Forecasting Enters the Scaling Era

Toto 2.0 is a family of five open-weights time series foundation models scaling from 4M to 2.5B parameters, achieving state-of-the-art on BOOM, GIFT-Eval, and the contamination-resistant TIME benchmarks. A single training recipe drives consistent forecast quality improvements across the full parameter range, validating scaling laws for time series. Teams running observability or general forecasting pipelines should evaluate these as drop-in replacements for task-specific models.