Recent Grok 4.7 releases have lifted xAI’s large language model into the low-to-mid 40s on Humanity’s Last Exam leaderboards, including around 43% on Artificial Analysis evaluations and lower no-tools HLE-Diamond runs, anchoring trader consensus around the 42%+ threshold. Frontier competitors such as Claude Opus 5.5 and Gemini variants lead at 55-61% on comparable splits of the 1,000-question reasoning-and-knowledge benchmark, while tool-augmented runs push select models above 65%. xAI’s ongoing Colossus expansion and announced larger training runs could deliver incremental gains before year-end, yet the two-month window leaves higher bars like 50%+ more contested amid rapid capability shifts across labs.