Opus 5.5 beats GPT-6 Sol in 7 of 10 real-world tests; Sol wins code repair

TawohAwa · x · 2026-09-23

Blogger nateherk ran Claude Opus 5.5 and GPT-6 Sol through 10 real tasks: websites, video editing, business deliverables, games, research, code repair, and browser use. Score: 7 wins for Opus, 1 for Sol, 2 unscorable.

Original post →

More from Models

Models channel →