Can PyMC be used for reversible-jump MCMC?

Note, however, that this does not yet work for posterior inference (i.e. sampling). The reason is that the trace backend (arviz.InferenceData ) as well as samplers in this case also must support changing dimensionality (like reversible-jump MCMC). There are plans to add this.

is these update now?

Thanks

No, nobody ever managed to provide a proof of concept of how the algorithm should be implemented. Some exploration was done sometime ago: rjmcmc stepper and example test case by TeaWolf · Pull Request #20 · pymc-devs/pymc-experimental · GitHub but nothing came out of it