OpenAI’s GPT-6 Astra and related variants trail Anthropic’s Claude Opus 5.5 and Fable 5.1 on October 2026 HLE leaderboards, where the top Claude models reach 59–65% while OpenAI’s strongest reported scores sit near 54.7–58.7%. This gap reflects Anthropic’s edge in frontier reasoning benchmarks, though OpenAI has demonstrated sharper gains on tool-augmented variants like HLE-Diamond (82.9% for GPT-6 Astra). Traders are pricing high odds for OpenAI clearing 55% by year-end but see meaningful uncertainty above 60%, driven by the pace of upcoming model releases, post-training optimizations, and potential benchmark updates before December 31.