GluonTS
A Python library for deep learning-based probabilistic time series forecasting and modeling built on PyTorch.
Build, train, and evaluate deep learning forecasting models with prediction intervals in Python.
GluonTS is a Python toolkit designed for probabilistic time series forecasting using PyTorch. It provides pre-built neural network architectures like DeepAR and utilities for evaluating forecast distributions. You should use it if you need uncertainty estimates alongside time series predictions rather than point estimates.
What it is
The package allows you to load time series data via Pandas, split datasets for evaluation, and train neural forecasting models. It outputs predictions as full probability distributions with customizable confidence intervals. It also integrates with external pretrained foundation models like Chronos for zero-shot time series tasks.
Who it's for
It suits data scientists, machine learning engineers, and quantitative analysts building complex forecasting pipelines. It is not designed for non-programmers who need a graphical user interface or simple spreadsheets for basic linear projections.
Key features
Probabilistic Forecast Distributions
Generates full predictive probability distributions and shaded prediction intervals rather than single point forecasts.
PyTorch Neural Estimators
Includes neural network models like DeepAR built directly on top of the PyTorch framework.
Dataset Splitting Utilities
Provides abstractions like PandasDataset and test-generator utilities for rolling-window time series evaluation.
Pretrained Model Support
Supports zero-shot forecasting integration with pretrained foundation models such as Chronos.
Up and running in 3 steps
Our honest verdict
What's great
- Provides native support for probabilistic output distributions and prediction intervals.
- Includes pre-built neural forecasting algorithms like DeepAR out of the box.
- Integrates smoothly with Pandas DataFrames and standard Python data science tools.
- Offers built-in abstractions for generating test instances and rolling-window evaluations.
Where it falls short
- Requires Python programming expertise and understanding of deep learning concepts.
- Training complex neural time series models on large datasets requires GPU compute resources.
- Lacks a graphical user interface for non-technical business users.
Pricing & license
GluonTS is free to use under the Apache-2.0 open-source license. The primary cost involved is provisioning the CPU or GPU infrastructure required to train deep learning models on your data.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What Python versions does GluonTS support?
GluonTS requires Python versions 3.10 through 3.14.
Which deep learning framework does GluonTS use?
GluonTS is primarily built on top of PyTorch.
Does GluonTS provide point estimates or probability distributions?
It produces probabilistic forecasts, allowing you to calculate prediction intervals such as 50% and 90% bounds.
