BlackJAX is a Python library designed for Bayesian computation, focusing on ease of use, speed, and modularity. It implements sampling and variational inference algorithms using a functional approach, allowing users to build and experiment with new algorithms by composing basic components. BlackJAX is built on JAX, enabling efficient execution on CPUs, GPUs, and TPUs. The library integrates well with probabilistic programming languages (PPLs) by working directly with the target log density function. It supports various sampling algorithms such as MCMC, SMC, and Stochastic Gradient MCMC (SGMCMC), as well as approximate inference methods like VI. BlackJAX provides a low-level API for users who need more control over the implementation of complex algorithms. The library is designed to be composable, allowing users to build custom algorithms by combining existing components. BlackJAX has been widely adopted in research, education, and practical applications, contributing to the development of new Bayesian sampling methods and enhancing the transparency and reproducibility of Bayesian inference. Future plans include expanding the library's methods, improving documentation, and enhancing performance diagnostics.BlackJAX is a Python library designed for Bayesian computation, focusing on ease of use, speed, and modularity. It implements sampling and variational inference algorithms using a functional approach, allowing users to build and experiment with new algorithms by composing basic components. BlackJAX is built on JAX, enabling efficient execution on CPUs, GPUs, and TPUs. The library integrates well with probabilistic programming languages (PPLs) by working directly with the target log density function. It supports various sampling algorithms such as MCMC, SMC, and Stochastic Gradient MCMC (SGMCMC), as well as approximate inference methods like VI. BlackJAX provides a low-level API for users who need more control over the implementation of complex algorithms. The library is designed to be composable, allowing users to build custom algorithms by combining existing components. BlackJAX has been widely adopted in research, education, and practical applications, contributing to the development of new Bayesian sampling methods and enhancing the transparency and reproducibility of Bayesian inference. Future plans include expanding the library's methods, improving documentation, and enhancing performance diagnostics.