Recent releases of OpenAI’s GPT-5.5 variants in late April have lifted the company’s best models to the low-to-mid 40s on Humanity’s Last Exam, a 2,500-question benchmark testing graduate-level reasoning across dozens of fields. Google’s Gemini 3.1 Pro Preview currently leads the public leaderboards with scores near 45 percent, while OpenAI models trail by a few points depending on the exact configuration and evaluation run. Traders are watching for any June model updates, internal scaling improvements, or new reasoning techniques that could close the gap before the June 30 cutoff. Historical patterns show frontier labs can deliver rapid benchmark gains between major releases, though timelines often slip and exact thresholds for market resolution remain sensitive to leaderboard updates.