All About AI runs a second weekly comparison between GPT 5.6 and Claude Opus 5 using questions drawn from prediction markets. The test covers first-week album sales, the United States unemployment rate and the following day's highest temperature in London.
Both models receive the same question options and resolution rules, while current market prices are hidden to reduce contamination from the crowd forecast. Each model is run separately at high reasoning effort, and the first model's work is removed before the second run.
The two systems select the same unemployment-rate and London-temperature outcomes but choose different ranges for album sales. The experiment therefore shows substantial agreement on two questions while leaving one observable point of divergence for later scoring.
Watch on YouTube
