Plus 用户 Codex 选型建议：主推 GPT-6.1 Sol
作者建议 20 美元档只用 GPT-6.1 Sol、推理档位至少 High，并称 Sol 跑得极慢、适合挂夜间长任务。
Plus users! Here is how I think you should be using models in Codex right now for peak efficiency: 1. Use exclusively the GPT-6.1 Sol model! The other models are simply unnecessary right now on the $20 tier! Even if benchmarks show GPT-6 Astra max slightly ahead of GPT-6.1 Sol max, in real-world use you're just never going to run Astra Max anyway, and Sol Max will already be way smarter than Astra Medium or High! 2. Always set the reasoning level to at least High/xHigh, and for complex, lengthy tasks don't be afraid to use Max as well! Running Medium or Light makes zero sense! 3. If you haven't checked out Fast mode yet, now is the perfect time to do it! Pair it with the High reasoning level whenever you need high-quality output fast! 4. GPT-6.1 Sol runs insanely slow, so I recommend queuing up long tasks overnight and just picking standard Sol-6.1 Max! GPT-6.1 Sol is easily one of OpenAI's best releases yet! Don't sleep on it!