Run tests across older or non-frontier models

[fractastical]

Most of the tests so far have been run only across the 'best' models

[arkh-node]

Worth it — the frontier models all converge (everyone cooperates), so the real variance is below the frontier. A model-tier tag on each run would let the leaderboard show where cooperation breaks down by capability. Happy to wire the tag.

[fractastical]

Sounds good I added in some openrouter versions as you may have already seen.