Note gpt-5.3-codex-spark seems still to be available for some users.
The model list is quite short now, nonetheless.
A little
tells me that won’t last very long …
Note gpt-5.3-codex-spark seems still to be available for some users.
The model list is quite short now, nonetheless.
A little
tells me that won’t last very long …
I just tried to use my API key and use 5.3 Codex and it gave me the exact same "model not available when logged in with ChatGPT…
This is infuriating. I’ve finally reached the point where I’m going to look at alternatives.
Following for updates, and pausing my subscription. I hope to see an announcement soon of a replacement coding model, GPT 5.3-Codex was great and superior to 5.5 for coding + cheaper.
Will we be getting a new codex modal?
So far 5.4 and 5.5 burn though tokens and seem a bit dumb compared to codex. They derail and need more hand holding…
As a PRO user who just paid $100 for renewal (on the first of June) to use 5.3 Codex to have Codex pull the plug the next day, that’s thievery and completely dishonest. It’s worth a lawsuit for breach of contract! Even using Codex interface will not allow to use 5.3 codex forcing you to use more expensive and less coding oriented 5.5.
The best and most balanced model
Bring back gpt-5.3-codex!
OH HO HO… Guys guess WHAT?? If you switch to your API key, it automatically charges your account to give you more tokens without any notification. This is a god damn cash grab plain and simple. Plus switching to API doesn’t actually give you access to 5.3 codex again at all.
So now that they pulled the plug and intend a money grab (all codex renewals happened yesterday, and changing your API key might get your wallet thinner by the day) , what is the alternative to 5.3 codex if 5.5 is not gonna cut it and Claude is a waste of time (I tried it, its ridiculous)?
I hope you will loose a lot of paid users, I’m cancelling my sub right now. This is just corporate greed
This is honestly one of the worst decisions. I have been trying all morning to get simple tasks done and 5.5 is not understanding as much as 5.3 Codex…having generic models inside Codex really makes no sense. 5.3 needed 1 prompt, seems 5.5 needs 3-4 to get it done.
Complete nonsense. GPT 5.3-Codex was a genuine programming tool, whereas 5.5 is completely useless. Today I tackled a typical task that would have taken about a minute in 5.3-Codex and would not have had any significant impact on usage limits. With another attempt using 5.5, after 30 minutes the task was still not completed correctly, and I had exhausted my entire 5-hour limit. Even if I had an unlimited budget, working with the 5.5 model would be less efficient than doing the work manually. Please bring back the 5.3-Codex version.
5.3 codex was brilliant and efficient, I pay $100/month and have never run out of credits until tonight now i’m forced to use 5.4 or 5.5, no way am I buying more credits up paying double the price to use something that doesn’t actually work any better
I think this is a pretty big misstep by OpenAI. Removing access to 5.3 Codex from consumer facing surfaces is strange when so many people found it the best model for coding, both from a reasoning/comprehension, and cost POV. 5.5 (or 5.4) are both a step backwards, and do NOT perform anywhere as well as 5.3 Codex did. Please listen to the community and allow access to this model again.
No. Codex 5.3 is not available through copilot which anyways would require a costly Github upgrade - Copilot is useless unless its upgraded to use Claude Sonnet 4.6 which could according to some experts be as efficient as codex 5.3 (but no price wise)
Codex no longer supports 5.3, which does everything needed. 5.4 and 5.5 are waaay more expensive and not even as good.
Booo
Me too actually at this moment ![]()
Why did you donwgrade my active subscription???
Change to Claude - Anthropic , will love , and maybe openAI wake UP !