Grok 4.7 vision testing: better detection than 4.6, still far behind leaders

hexiang · x · 2026-09-23

Quoted testing shows Grok 4.7's vision is improving: better detection than 4.6, more accurate bounding boxes across domains, faster and cheaper at high effort—but still far behind leaders, struggles with crowded scenes, and shows some regressions. xAI responded that they will keep improving.

Original post →

More from Models

Models channel →