OpenAI Codex 5.4
Thus far I like it better than Codex 5.3
I’m using Codex 5.4 with high reasoning in terminal. OpenAI released this model yesterday. It is currently fixing bugs for code written by 5.3 earlier last week for a new product feature. 5.4 found where the code was brittle in approx. 30 seconds. After 3 hours using Codex 5.4 I am impressed.
Codex 5.4 has a 1 million token context window, up from approx. 200,000 with codex 5.3.