**Grok models currently trail frontier performance on Humanity’s Last Exam (HLE), a 2,500-question expert-authored benchmark testing graduate-level knowledge and multi-step reasoning across math, sciences, and humanities.** Recent October 2026 leaderboards show Grok 4.7 at approximately 43% and earlier variants in the low-to-mid 40s, well behind Anthropic’s Claude Opus 5.5 (61%+) and related variants that lead multiple evaluations. The remaining time until the December 31, 2026 resolution deadline is short, so any new xAI model releases, scaling improvements, or tool-augmented setups would need to deliver substantial gains quickly to close the gap. Market-implied odds reflect trader assessment of xAI’s development velocity relative to competitors amid rapid benchmark progress by other labs.