Back to Gap Map Infrastructure / Classic ML & computer vision

H2O-3

open core / Overall score: 3.8

H2O.ai's open-source, in-memory, distributed machine-learning platform, running on the JVM with Python and R clients. It offers GLM, GAM, gradient boosting, XGBoost, random forests, deep learning, SVMs, stacked ensembles, isolation forests, k-means and PCA, and H2O AutoML, which trains, tunes and stacks a set of models within a time or model budget. Models export as MOJO and POJO artifacts for deployment. H2O.ai sells a commercial H2O-3 Secure edition alongside it.

Openness

4 medium confidence

4.0

license: Apache-2.0(OSI)
source: public(github.com/h2oai/h2o-3)
core features withheld: yes — H2O-3 Secure, under a commercial license, adds early-access algorithms, enterprise Hadoop and Kubernetes packages and production MOJO scoring

The open H2O-3 is Apache-2.0 and includes every algorithm and AutoML, but H2O.ai sells H2O-3 Secure under a commercial license that adds early access to new algorithms, Hadoop and Kubernetes enterprise packages, production scoring for Hadoop, Spark, Teradata and Oracle, and security and compliance features. Part of what H2O-3 users run in production is therefore held back from the open release.

https://h2o.ai/h2o-3/oss-vs-secure/ recorded 2026-09-27
Comparison table "H2O-3 OSS Apache 2.0 | H2O-3 Secure Commercial License": both have all algorithms, H2O AutoML, distributed training and scoring; Secure only has "Early access to latest algorithms and improvements", "Hadoop and Kubernetes enterprise packages", "Production MOJO model scoring for Hadoop, Spark, Teradata, Oracle", "Managed distributed XGBoost service", "Automated model documentation (H2O Auto Doc)" and enterprise SSO.
https://h2o.ai/products/h2o/ recorded 2026-09-27
H2O product page: "H2O-3 OSS is free and Apache-licensed, designed for self-managed and experimental workflows. Running H2O-3 in production environments? Upgrade to H2O-3 Secure: same code, APIs and pipelines."
https://pypi.org/pypi/h2o/json recorded 2026-09-27
PyPI record for h2o 3.46.0.12, uploaded 2026-08-12, license "Apache v2", Homepage github.com/h2oai/h2o-3.
https://raw.githubusercontent.com/h2oai/h2o-3/master/LICENSE recorded 2026-09-27
LICENSE is the Apache License, Version 2.0, "Copyright 2014-2021 H2O.ai, Inc.".
https://ungh.cc/repos/h2oai/h2o-3/files/master recorded 2026-09-27
Repository tree of about 8,500 paths with no ee, enterprise, commercial or pro directory; the Secure additions are not in it.

Adoption

3 medium confidence

3.0

Measured on monthly PyPI downloads of the h2o Python client. The R package on CRAN and direct JVM use are not counted.

https://pypistats.org/api/packages/h2o/recent recorded 2026-09-27
last_month 156,866 downloads of h2o

Capability

5 high confidence

5.0

H2O AutoML takes the user's table and a time or model budget and returns a leaderboard topped by stacked ensembles, with no model choice on the user's side, which is the automation AutoGluon offers. Unlike AutoGluon it bundles no pretrained tabular models.

https://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html recorded 2026-09-27
AutoML FAQ: "The current version of AutoML trains and cross-validates the following algorithms: three pre-specified XGBoost GBM (Gradient Boosting Machine) models, a fixed grid of GLMs, a default Random Forest (DRF), five pre-specified H2O GBMs, a near-default Deep Neural Net, an Extremely Randomized Forest (XRT), a random grid of XGBoost GBMs, a random grid of H2O GBMs, and a random grid of Deep Neural Nets", with Stacked Ensemble models built on them.
https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science.html recorded 2026-09-27
Algorithms page: Supervised - H2O AutoML, CoxPH, Deep Learning, DRF, GLM, Isotonic Regression, GAM, HGLM, GBM, Naive Bayes, RuleFit, Decision Tree, AdaBoost, Stacked Ensembles, SVM, Uplift DRF, XGBoost; Unsupervised - Aggregator, GLRM, Isolation Forest, Extended Isolation Forest, K-Means, PCA.

Verified 2026-09-27