“According to Artificial Analysis benchmarks, Claude Opus 4.8 uses 80-90% more tokens than GPT-5.5 to achieve a comparable intelligence score, making it significantly less token-efficient.”