aiqbench analysis
Toto 2.0 demonstrates that time series foundation models achieve reliable performance scaling from 4M to 2.5B parameters using a single training recipe, establishing new state-of-the-art results across observability, general-purpose, and contamination-resistant benchmarks. The release of five open-weight checkpoints under Apache 2.0, alongside detailed architectural insights and a hyperparameter transfer pipeline, provides a practical foundation for building robust forecasting systems. This work validates the efficacy of scaling laws in temporal data, offering a standardized approach for future model development in this domain.
Toto 2.0: Time Series Forecasting Enters the Scaling Era
arXiv:2605.20119v2 Announce Type: replace-cross Abstract: We show that time series foundation models scale: a single training recipe produces reliable forecast-quality improvements from 4M to 2.5B parameters. We release Toto 2.0, a family of five open-weights forecasting models trained under this recipe. The Toto 2.0 family sets a new state of the art on three forecasting benchmarks: BOOM, our observability benchmark; GIFT-Eval, the standard general-purpose benchmark; and the recent contamination-resistant TIME benchmark. This report describes our experimental results and details the design decisions behind Toto 2.0: its architecture and training recipe, training data, and the u-muP hyperparameter transfer pipeline. All five base checkpoints are released under Apache 2.0.
Abstract:We show that time series foundation models scale: a single training recipe produces reliable forecast-quality improvements from 4M to 2.5B parameters. We release Toto 2.0, a family of five open-weights forecasting models trained under this recipe. The Toto 2.0 family sets a new state of the art on three forecasting benchmarks: BOOM, our observability benchmark; GIFT-Eval, the standard general-purpose benchmark; and the recent contamination-resistant TIME benchmark. This report describes our experimental results and details the design decisions behind Toto 2.0: its architecture and training recipe, training data, and the u-muP hyperparameter transfer pipeline. All five base checkpoints are released under Apache 2.0.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2605.20119 [cs.LG]
(or arXiv:2605.20119v2 [cs.LG] for this version)
Full text under the publisher's open licence · 116 words
