They have the preferred model for almost anything that reasoning isn't a perfect fit for, and even on those tasks only O1 wins consistently, for >10x pricing an

They have the preferred model for almost anything that reasoning isn't a perfect fit for, and even on those tasks only O1 wins consistently, for >10x pricing and latency. Even for reasoning-beneficial tasks you can just CoT Sonnet and it works well, just yesterday I had a task's performance strongly improved by it.