Recent model releases have driven OpenAI's positioning on Humanity’s Last Exam, a 2,500-question benchmark testing frontier-level reasoning across math, science, and humanities. GPT-5.5 variants released in April 2026 reached 43–44.3% accuracy, narrowing the gap with Google's leading Gemini 3.1 Pro Preview at 44.7%, while earlier GPT-5.4 models sat near 41.6%. Trader sentiment reflects expectations that further scaling, specialized reasoning enhancements, or new variants could push an OpenAI GPT model higher before the June 30 resolution window, though historical patterns show benchmark gains often plateau without major architectural shifts. Key catalysts include any additional OpenAI announcements or fine-tuning updates in the coming weeks.