Member-only story
Amazing. Well, it depends on who you ask. In my opinion, GPT-5.6 is my favorite release yet.
OpenAI also changed the part that makes these models usable for serious coding work. At the time of writing, the five-hour Codex cap has disappeared for my account. I see a weekly limit instead. I can run hard coding sessions until that weekly reset rather than watching a short five-hour clock disappear halfway through a job.
That does not mean every plan has unlimited Codex usage. It is an account-level observation, and limits can change. But it is a big practical change for me. The five-hour wall was the annoying part of using Codex heavily. Weekly limits are still limits, but they fit long coding work far better, espically with the free resets OpenAI gives, I had 4, so there is a lot of coding to do :D.
I am not excited about this because OpenAI added one more expensive flagship. I am excited because Sol, Terra, and Luna give you three serious ways to spend less than the GPT-5.5 era while improving the family’s public performance story. Sol reaches 59 on the Artificial Analysis Intelligence Index, one point behind Claude Fable 5. Terra reaches 55 while cutting GPT-5.5’s published token rates in half. Luna reaches 51 at $1 per million input tokens and $6 per million output tokens.