PyMC is the leading Python library for Bayesian modeling and probabilistic programming, letting data scientists express statistical models in code and fit them with state-of-the-art inference. Developed by the PyMC team with a large open-source community, it has nearly 10,000 stars on GitHub and remains in active development in 2026.
PyMC turns model specification into a readable Python API: define priors and likelihoods, and the library compiles them onto the PyTensor backend for automatic differentiation and fast execution. Inference options include the No-U-Turn Sampler (NUTS) and other MCMC methods, variational inference for large problems, and specialized samplers for discrete and time-series models. Tight integration with ArviZ provides posterior diagnostics, model comparison, and publication-quality plots, while the ecosystem includes PyMC-Marketing and other domain libraries built on top of the core. The result is a workflow where uncertainty is a first-class output: every estimate comes with credible intervals, and model checking is part of the normal process rather than an afterthought bolted onto a point forecast.
Free. PyMC is a fully open-source community project; the library, documentation, and examples cost nothing, and models run entirely on your own infrastructure.
Data scientists, statisticians, and researchers who need principled uncertainty quantification โ forecasting, A/B analysis, marketing mix modeling, and scientific inference where credible intervals matter as much as point estimates. It is also a strong teaching tool for Bayesian statistics, since models are written as explicit, inspectable code rather than hidden inside a black-box service.