Sonnet 5.5 has the same problem as Sonnet 5
retained text
Don't understand who Sonnet 5.5 is for. Low, medium, and high effort, GPT-6 Sol performs better while using fewer output tokens. Sonnet 5.5 needs 193k output tokens per task just to get close to Opus, 60% more than Opus 5.5 Max and 7x GPT-6 Astra Max. Costs $7.60 per task compared to $5.98 for Opus 5.5 Max. If I care about efficiency, Sol is better. If I want the best Claude, Opus scores higher while literally costing less per task at max.
document fetched 2026-09-30 00:23 · canonical https://reddit.com/r/ClaudeAI/comments/1wt3o8m/sonnet_55_has_the_same_problem_as_sonnet_5