I’m sorry codex but I’m going back to Claude code.
I’m genuinely surprised by how fast and accurate GPT-5.4 (Codex) is compared to Opus 4.6.
I still have access to my Claude subscription, so I’ve been running the same prompts on both... and Codex consistently performs faster and more accurately.
What’s even crazier: Claude was on the max plan, while Codex was just on ChatGPT Plus.
Rate limits are also better on Codex.
Honestly, for coding, Codex feels like the way to go.