LightGBM
LightGBMGradient-boosted decision tree library built for speed and memory efficiency, with histogram-based tree learning, leaf-wise growth, GPU training and distributed learning. It handles regression, classification and learning-to-rank from Python, R, C and a command line. Created at Microsoft, it moved in March 2026 to the lightgbm-org account under the same maintainers.
Openness
5 high confidence- license
- MIT(OSI)
- source
- public(github.com/lightgbm-org/LightGBM)
- core features withheld
- no — no paid tier or commercial edition found
MIT-licensed and built from the public repository, which moved from Microsoft's GitHub account to lightgbm-org in March 2026 with the same maintainers. Nobody sells an edition of it.
- https://raw.githubusercontent.com/lightgbm-org/LightGBM/main/LICENSE recorded 2026-09-27
LICENSE is "The MIT License (MIT)", "Copyright (c) Microsoft Corporation" and "Copyright (c) The LightGBM developers".
- https://raw.githubusercontent.com/lightgbm-org/LightGBM/main/README.md recorded 2026-09-27
README: "This project moved from `Microsoft/LightGBM` to `lightgbm-org/LightGBM` in March 2026. This repository is still the official LightGBM source code, managed by the same maintainers (including the creator of LightGBM)."; "licensed under the terms of the MIT license"; no paid tier mentioned.
- https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lightgbm-org/LightGBM recorded 2026-09-27
Repository record for lightgbm-org/LightGBM, not archived and not a fork, license mit, last pushed 2026-09-24.
- https://ungh.cc/repos/lightgbm-org/LightGBM/files/main recorded 2026-09-27
Repository tree of 566 paths with no ee, enterprise, commercial or pro directory.
Adoption
5 high confidenceMeasured on monthly PyPI downloads of the lightgbm package; the R package and the command-line build are not counted.
- https://pypistats.org/api/packages/lightgbm/recent recorded 2026-09-27
last_month 18,191,249 downloads of lightgbm
Capability
2 high confidenceLightGBM fits one kind of model, boosted trees, as XGBoost does, and competes with it on speed and memory rather than on range. Choosing between it and other learners, and tuning it, is left to other tools; its own README points to FLAML and Optuna for that.
- https://lightgbm.readthedocs.io/en/latest/Features.html recorded 2026-09-27
Features page: histogram-based algorithm, leaf-wise tree growth, DART, GPU and distributed learning; "LightGBM supports the following applications: regression ... binary classification ... multi classification ... cross-entropy ... LambdaRank".
- https://lightgbm.readthedocs.io/en/latest/Python-API.html recorded 2026-09-27
Python API lists LGBMModel, LGBMClassifier, LGBMRegressor, LGBMRanker and their Dask variants, all boosted-tree estimators.
Verified 2026-09-27