GPT-6.2 Leads 4,700-Vote Blind Marketing Benchmark, But Top Models Statistically Tied

A large-scale blind marketing benchmark with 35 real brands and over 4,700 votes found GPT-6.2 Sol beating Claude Opus 5.5 with a 58% win rate at just $0.04 per task, though the three top models were statistically tied overall.

2026-11-21 ~ 2026-11-21 · 2 related posts