Awesome PPL

A curated list of Probabilistic Programming Languages (PPLs), inference libraries, and learning resources.

Probabilistic programming integrates probabilistic models and code into a unified paradigm.
Instead of coding “how” to compute, you specify what your model is, and the PPL system automatically performs inference — from Bayesian statistical models to deep probabilistic programs.

---

Contents

• Python
• Julia
• R
• Scala
• JavaScript
• Logic & Specialized
• Contributing

---

Python

• PyMC – Declarative, graph-based PPL for applied Bayesian statistics. Supports NUTS, ADVI, SMC. Tutorial: PyMC Examples
• Pyro – Universal (generative) PPL built on PyTorch. Strong in deep probabilistic modeling, VI, and HMC/NUTS.
  Tutorial: Pyro Tutorials
• NumPyro – JAX-based Pyro, offering major speedups via JIT compilation.
  Tutorial: NumPyro Tutorials
• TensorFlow Probability – Library of probabilistic building blocks in TensorFlow/JAX. Supports HMC, VI, and optimizers.
  Tutorial: TFP Examples
• Stan (via CmdStanPy/PyStan) – Declarative, static-graph language with gold-standard NUTS implementation.
  Tutorial: Stan User’s Guide
• Bean Machine – Meta’s research PPL for programmable inference using PyTorch.
  Tutorial: Bean Machine Tutorials
• Edward2 – A lightweight, low-level probabilistic programming library built on TensorFlow, JAX, or NumPy, emphasizing flexibility via a single abstraction (“random variable”) and tracing-based model manipulation :contentReference[oaicite:0]{index=0}.
  Tutorial: Edward Tutorials :contentReference[oaicite:1]{index=1}
• GenJAX – JAX-based implementation of Gen with programmable variational inference.
  Tutorial: GenJAX Documentation
• BlackJAX – Sampling algorithms for JAX (HMC, NUTS, Riemannian HMC, SMC) designed for composability with other JAX-based PPLs.
  Tutorial: BlackJAX Examples

Julia

• Turing.jl – General-purpose, high-performance PPL supporting NUTS, HMC, PMCMC, Gibbs, VI.
  Tutorial: Turing.jl Getting Started
• Gen.jl – Research-oriented PPL with programmable inference, hybrid MCMC/VI/SMC algorithms.
  Tutorial: Gen.jl Documentation

R

• Stan (via rstan, cmdstanr) – Seamless integration with R workflows.
  Tutorial: Stan User’s Guide
• greta – Write models in R syntax, backed by TensorFlow for CPU/GPU execution.
  Tutorial: greta Getting Started
• JAGS/BUGS – Pioneering declarative PPLs using Gibbs sampling.
  Tutorial: JAGS Documentation

Scala

• Figaro – Expressive Scala PPL with broad algorithm support (MCMC, SMC, exact inference).
  Tutorial: Figaro Guide
• Rainier – High-performance, static-graph Scala PPL inspired by Stan/PyMC.
  Tutorial: Rainier Examples

JavaScript

• WebPPL – Universal PPL for the web. Runs in-browser or Node.js, popular in cognitive science.
  Tutorial: WebPPL Web Book

Logic & Specialized

• ProbLog – Probabilistic logic programming over large, uncertain knowledge bases.
  Tutorial: ProbLog Tutorials
• Dice – High-speed exact inference for discrete probabilistic programs via WMC.
  Tutorial: Dice Documentation

---

Contributing

Contributions are welcome! Please ensure:

• Each entry has a link, short description, and follows the existing format.
• Include an official tutorial link when available.
• Keep descriptions objective and concise.

---

Last updated: 2025-08-14