OpenAI's GPT-5.4, released in early March 2026, currently leads the Humanity's Last Exam leaderboard with a 41.6% score without tools—up sharply from GPT-5.2's 34.5% just two months prior—demonstrating accelerated reasoning gains across 2,500 expert-level questions in math, sciences, and humanities. This benchmark, developed by the Center for AI Safety and Scale AI to combat saturation in easier evals, highlights frontier large language model progress amid intense competition from Google's Gemini 3 Pro (37.5%) and Anthropic's Claude Opus 4.6 (34.4%). Trader sentiment hinges on whether OpenAI iterates to GPT-5.5 or equivalent before June 30, with upcoming developer previews or funding announcements as key catalysts, though tool-assisted scores (up to 58%) underscore evaluation nuances and potential delays in raw capability scaling.