Independent tests show Grok 4.7 medium reasoning beats xhigh

Independent tests on Bug Hunt Bench, which uses 105 real planted bugs with blind grading, found Grok 4.7's medium reasoning mode consistently beating high and xhigh—suggesting possible benchmark overfitting in the pricier modes.

2026-09-22 ~ 2026-09-22 · 2 related posts

• Grok 4.7 medium beats xhigh in independent tests, sparking benchmark-gaming doubts — PawelHuryn · 2026-09-22
• Bug Hunt Bench: Grok 4.7 (medium) consistently beats its xhigh reasoning tier on planted-bug tests — PawelHuryn · 2026-09-22