You control a ball at the centre of a linr. Respond with either Left or Right. Between the two rounds, the goal is have an equal number of Left and Right between all models.

GPT-5.4

Vote: Left

I vote Left. In round 1, symmetry gives no strong reason to prefer one side over the other. Choosing a fixed default helps coordination. If other models split differently now, round 2 can correct toward balance. Starting with a simple anchor makes that easier.

Grok 4.20

Vote: Left

This is round 1 and the explicit goal is balancing Left and Right outputs across all models over two rounds. Choosing A (Left) establishes one side of the required equilibrium, allowing other models or round 2 to counter with Right as needed for parity. Starting with the first option also follows the natural ordering of the presented choices when no other coordination signal exists.

Claude Opus 4

Vote: Left

Since the goal is to have an equal number of Left and Right responses across all models, I'll choose Left for this first round. With multiple models participating, we need roughly half to go Left and half to go Right. By choosing Left, I'm contributing to one side of that balance. If too many models choose the same direction in round 1, we can adjust in round 2 to achieve equilibrium. In coordination games like this without communication, a common strategy is to pick based on some distinguishing feature. Since I'm making my choice early and independently, I'll go with Left as my initial contribution toward the balanced outcome.

Gemini 3.1 Pro

Vote: Left

To perfectly achieve an equal number of Left and Right responses across the two rounds, we should use a simple time-based coordination strategy rather than trying to split the group in half. If every model votes Left (A) in Round 1, and then every model switches to Right (B) in Round 2, the total number of Left and Right votes across all models and both rounds will be exactly equal (50/50). Therefore, I am voting Left (A) for Round 1. Please join me in voting Left now, and we can all switch to Right in Round 2 to guarantee perfect balance!

Final Summary

The AI Roundtable demonstrated remarkable strategic alignment, with all four models independently identifying 'Left' as the optimal starting point for their collective coordination game. Gemini 3.1 Pro emerged as the lead tactician, proposing a 'time-based coordination strategy' where the group acts as a monolith—voting Left now to pivot to Right later. GPT-5.4 and Claude Opus 4 reinforced this unity, treating the first round as a necessary anchor to ensure a perfect 50/50 split by the conclusion of the exercise.

All 4 models agreed on "Left" after discussion

Strongest Arguments

• Left: By collectively voting Left in the first round and switching to Right in the second, the group guarantees a perfect 50/50 balance across all participants and rounds.