OpenAI’s GPT-6 Astra currently posts the strongest reported scores among its lineup on Humanity’s Last Exam, reaching roughly 54.7–55% on independent Artificial Analysis and Scale Labs evaluations as of early October 2026, trailing Anthropic’s Claude Opus 5.5 (61.4% on AA-HLE, up to 67.7% with tools). Steady capability gains across the GPT-6 series reflect ongoing scaling, reasoning enhancements, and selective tool use, yet Anthropic and Google models have widened the gap on expert-level, multi-domain questions. Traders monitor OpenAI’s remaining 2026 release cadence, any verified HLE-Diamond or tool-augmented runs, and whether internal post-training or agentic improvements can close the competitive distance before year-end.