OpenAI’s GPT-6 Astra currently posts the strongest verified scores among its models on several HLE leaderboards, including 54.8% on the Scale full set and 59.9% on the HLE-Diamond subset without tools, though Anthropic’s Claude Fable 5.1 and Opus 5.5 lead other evaluations at 59–65% depending on the question subset, judge, and tool settings. Rapid progress across frontier labs—roughly 30 points gained industry-wide in the past year—has tightened the gap, with tool-augmented runs pushing OpenAI higher on Diamond variants while benchmark variations in image questions, calibration, and verification create resolution uncertainty. Traders are watching for any new OpenAI releases or system-card updates before year-end that could lift the highest 2026 score above key thresholds like 55% or 60% on the official Scale or equivalent leaderboard.