Rules

This market will resolve to “Yes” if Gemini 4, within 168 hours (7 days) after its first public release date that occurs after this market's creation, achieves a strictly higher numeric score than GPT-6 Astra on the ARC-AGI-3 benchmark, and both qualifying scores are publicly posted by credible sources within that 7-day window, before January 1, 2028 - 00:00 UTC.

Otherwise this market will resolve to “No.”

For the purposes of this market:

• “Public release” is the timestamp of the first public announcement or product/API availability from the model’s publisher (e.g., blog post, press release, product page) indicating general public access; the release must occur after market creation to be eligible.

• A “qualifying score” is the numeric score reported for a model on ARC-AGI-3 by the ARC-AGI-3 official leaderboard or, if no official leaderboard score is published, by the model publisher in an official benchmark report or model card; the score must reference the same ARC-AGI-3 dataset version and scoring metric.

• If multiple qualifying scores for a model are published within the 7-day window, the highest qualifying score published within that window will be used for comparison.

• Ties (identical numeric scores) result in a “No” resolution.

• If ARC-AGI-3 is permanently retired, materially redefined, or its official scoring metric changed before resolution such that scores are not comparable, the market may be cancelled; cancellation requires official benchmark maintainer confirmation.

• Only releases and benchmark scores published after market deployment are eligible for resolution.