On July 9, 2026, OpenAI released the GPT-5.6 series, consisting of 3 models and 15 reasoning variants. GPT-5.6 Sol, at max reasoning effort, is the standout model, averaging 13.33% on Public and 7.78% on Semi-Private, and is the first model to win an ARC-AGI-3 public game (ft09, 87%). Sol can read an unfamiliar scene correctly in the game's own vocabulary and treats a failed hypothesis as a reason to re-plan. Verified scores show Sol achieving 7.8% on ARC-AGI-3 at max reasoning, with scores decreasing to 0.3% at low reasoning. Terra and Luna scored 0.8% and 0.2% respectively on ARC-AGI-3 at max reasoning.