xAI’s Grok models currently trail on Humanity’s Last Exam, with Grok 4.7 and 4.6 posting 43.1% and 42.9% on leading verified leaderboards as of early October 2026, well behind Claude Opus 5.5 at 61.4% and other frontier systems. Recent releases emphasize larger parameter counts, expanded Colossus training clusters, and agentic capabilities, yet HLE scores have improved only incrementally. Traders are watching xAI’s aggressive cadence through Grok 4.8 and the anticipated Grok 5 before year-end, alongside any benchmark gains from additional compute or specialized fine-tuning, while noting that closing a roughly 18-point gap in under three months would require unusually rapid progress relative to historical model trajectories.