Comment by amanzi

Not surprised at all - just making the point that getting a model to run isn't that impressive if it runs at a few tokens per second.