Chronos-2 overview and key facts (October 2026)

Chronos-2 is an open time series foundation model from Amazon that produces forecasts without training on your data first. Data and planning teams use it to forecast demand, sales or capacity from historical series.

The model has 120 million parameters, handles contexts of up to 8,192 steps and predicts up to 1,024 steps ahead. It runs on GPUs and CPUs and is used through the chronos-forecasting Python package, SageMaker JumpStart or AutoGluon-Cloud. Hugging Face counted about 21.4 million downloads in the last 30 days as of September 28, 2026.

• Forecasts without trainingChronos-2 forecasts a new time series zero-shot, without fitting a model first. Teams get a baseline forecast quickly, even for series with little history.
• Uses related series and driversThe model supports multivariate forecasting and covariates such as prices or promotions. Forecasts can reflect factors that planners already know about.
• Shows forecast uncertaintyChronos-2 returns quantile forecasts over several steps. Planners see a range of likely outcomes and can plan stock or capacity with a safety margin.

• Small enough at 120 million parameters to run on a CPU, with GPU support for larger volumes.
• The model card reports more than 300 forecasts per second on a single A10G GPU, according to Amazon.

Chronos-2 is an open model from Amazon under Apache-2.0 and can be used worldwide without country ties.

You run the model yourself or on SageMaker, so plan hosting, monitoring and updates. Test forecast quality against your current method on your own history before relying on it. Data stays in your environment when you run the model locally.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews