OpenAI’s recent GPT-6 Astra release in early September 2026, along with GPT-5.4 variants, has driven trader sentiment by demonstrating strong gains on reasoning benchmarks, including HLE subsets with tools reaching the high 50s to low 80s on refined variants. However, the official HLE leaderboard as of mid-October shows Anthropic’s Claude Fable 5.1 and Opus 5 leading at 65% and 64.7%, while top OpenAI models like GPT-5.4 Pro sit at 58.7% and GPT-5.6 Sol near 54.5% on verified runs. Competitive dynamics hinge on whether OpenAI can close the gap through further scaling, tool integration, or new model drops before year-end, amid rapid frontier progress that has lifted top scores roughly 30 points over the past year. Key upcoming catalysts include potential GPT-6 iterations and official HLE updates that could shift the highest OpenAI mark toward or beyond the 60% threshold favored in current market-implied odds.