I tested 10 model/harness combinations on the same Three.js task
What happened
I've been testing a simple prompt with different model and harness combinations to work out which one produces best results. I do this in /goal mode.
Summary assembled by rule from the sources below