autogluon
Automates machine learning model training and ensembling for tabular, time series, and multimodal data in Python.
Train ensembled predictive models and foundation architectures across tabular, time series, and text data.
AutoGluon is an automated machine learning library designed for data scientists and ML engineers working in Python. It automatically finds, tunes, and ensembles traditional machine learning algorithms alongside foundation models for complex datasets. You should use it if you want high predictive performance without spending hours writing manual hyperparameter tuning pipelines.
What it is
AutoGluon takes raw datasets—including tabular tables, probabilistic time series, and multimodal inputs—and automatically runs model searches. It implements multi-layer stack ensembling and augmented distillation to combine diverse models into a unified predictor. Through task-specific modules like TabularPredictor, TimeSeriesPredictor, and MultiModalPredictor, it automates model training with minimal code.
Who it's for
It suits data scientists, predictive analysts, and machine learning engineers who need quick baselines or competitive benchmark models on structured data. It is less suitable for teams operating in severely resource-constrained embedded environments where running multi-model ensembles or deep learning foundation models is infeasible.
Key features
Tabular Data Automation
Trains and ensembles multiple classic machine learning models and deep architectures directly from structured data files.
Probabilistic Time Series Forecasting
Automates univariate and multivariate forecasting using multi-layer stack ensembles and dedicated time series foundation models.
Multimodal Model Training
Combines tabular inputs, unstructured text fields, and image data into unified predictive models using modern foundation architectures.
Model Distillation
Compresses complex multi-model stack ensembles into smaller, faster individual models via augmented distillation.
Up and running in 3 steps
Our honest verdict
What's great
- Automates model ensembling across tabular, time series, and text modalities in three lines of code.
- Supports multi-layer stack ensembling to maximize predictive accuracy over single-model approaches.
- Includes augmented distillation to generate lighter inference models from complex ensembles.
- Integrates deep foundation models like Chronos for time series forecasting alongside classic ML algorithms.
Where it falls short
- Training multi-layer stack ensembles requires high memory and computational resources, often needing dedicated GPU instances.
- Heavy dependencies can lead to complex environment configurations and large package installation footprints.
- Full ensemble models can introduce higher inference latency compared to single simple algorithms unless distilled.
Pricing & license
AutoGluon is completely free to use under the Apache-2.0 open-source licence. The primary operational cost comes from computing resources such as high-RAM CPUs or GPUs required to train and ensemble deep learning models.
Open remote AI-training roles
873 open remote AI-training roles, updated November 15, 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 AutoGluon support?
AutoGluon supports Python versions 3.10 through 3.13 across Linux, MacOS, and Windows operating systems.
Can AutoGluon handle time series forecasting?
Yes, it provides TimeSeriesPredictor for probabilistic time series forecasting using stack ensembling and zero-shot foundation models.
How does AutoGluon achieve high accuracy?
It trains multiple individual models and combines them using multi-layer stack ensembling rather than relying on a single best model.
