GitLab Transcend - Intelligent orchestration
Mini toolsChronos 2
By Amazon
Chronos-2 is Amazon’s next-generation time-series foundation model for universal forecasting. It is designed to work across univariate, multivariate, and covariate-informed tasks in zero-shot settings, using in-context learning rather than task-specific retraining. Amazon says the model jointly forecasts coevolving series, incorporates past-only and known-future covariates, and supports arbitrary numbers of targets and optional covariates. Its architecture uses patch embeddings plus alternating time-attention and group-attention layers to model interactions across series, and Amazon reports state-of-the-art results on fev-bench and first place among pretrained models on GIFT-Eval.
Text
Gen 7
Overview
Chronos-2 is Amazon’s universal time-series forecasting foundation model. It extends the earlier Chronos line beyond single-series prediction to handle univariate, multivariate, and covariate-informed forecasting in a zero-shot manner, using in-context learning to adapt to tasks with an arbitrary number of dimensions without extra training.
About Amazon
View Company ProfileTools using Chronos 2
No tools found for this model yet.
Last updated: April 8, 2026
