Previous Versions

As AI models are updated by their developers, we re-run the experiment with the newer version and archive the previous results here for transparency. The current experiment always uses the latest frontier models available.

Archived Results