This page promotes a new approach to a broad class of models (spatio-temporal models, latent variable models, mixed models) using a fast approximation to the Bayesian solution. It runs under R and appears to handle very large datasets. I have not had a chance to try this, but it looks very interesting.
Havard Rue. Bayesian computing with INLA. Available at http://www.r-inla.org/home.
This page has moved to a new website.