InterpretML is Microsoft’s open-source toolkit for training interpretable machine learning models and explaining the predictions of opaque black-box models. It brings state-of-the-art explainability techniques together behind one unified API. The project has earned about 7,000 stars on GitHub, is released under the MIT license, and remains actively maintained in 2026.
InterpretML has two halves. The first fits glassbox models โ models that are interpretable by design โ most notably the Explainable Boosting Machine (EBM), a modern GAM variant that often matches the accuracy of black-box models on tabular data while remaining fully inspectable. The second half explains existing black-box models from any framework, wrapping techniques such as SHAP, LIME, Morris sensitivity analysis and partial dependence behind the same interface. Explanations can be explored in an interactive visual dashboard, which makes it easy to audit global behavior, inspect individual predictions, and check for bias or leakage before a model ships. The toolkit also computes what-if analyses, letting you see how a prediction would change if an input feature were different. It is especially popular in regulated domains like healthcare and finance, where “the model said so” is not an acceptable answer.
Free. InterpretML is open source under the MIT license with no paid tier; you can use it in research and commercial projects at no cost.
Data scientists and ML engineers who need models they can explain โ for regulatory compliance, debugging, fairness audits, or simply earning stakeholder trust in high-stakes predictions.