Supposedly OpenAI has released the Codex models to use for free until December 11. I just ran out of credits on my Plus plan, and I’m getting the message that I have no credits left. My credits don’t renew until the 22nd :_(
Works for me too, currently! Just maxed out my $20 reload on Codex 5.1. Tried switching to Mini, but it still shows Codex Max (free until Dec 11). I already spent $20 a few days ago and another $20 today, and I hit the cap again. If I add another $20, will it automatically upgrade me to Pro+ for the rest of the month? Any insight would be appreciated.
It’s not even worth your time lol. It’s honestly pretty bad. Even ChatGPT-5.1-Thinking in the MAC OS app agreed. I had ChatGPT-5.1-Thinking provide a detailed prompt, and build the plan, and ChatGPT-Thinking described Codex Max’s plan as. “Yeah, that plan is still mostly vibes and marketing, not what you asked for. It’s better than the totally generic fluff, but it’s still failing the core requirement: it never proves it actually looked at your real src/server.ts or PLAN.md. It’s just restating the architecture we’ve already talked about, then sprinkling “implementation-ish” language on top.” rofl.
I recommended the new gpt-5.1-codex-max model in Cursor to my friend because the announcement said it is free to use until December 11th. However, when my friend tried to use it, it prompted him to upgrade to a Pro plan in order to access the model.
Is the model actually free, or is it only free for users who already have a Pro subscription? Could someone provide more details or clarify how the free-access period is supposed to work?
Yes, it shows me as free. The non-fast ones are slow and degrade the developer experience. Just asking a simple question about a 200-line file takes a lot of time. But guys, the xhigh-fast version gives the best answer at high speeds - I highly recommend using them while it’s free.
GPT 5.1 Codex family is pretty bad — it’s not even an option in agent mode. Even though it sounds strange, Auto mode works much better for me (when it uses Plan, it usually handles things quite well).
Honestly, the only model that solves things efficiently is Claude Sonnet/Opus 4.5. By far. (Maybe Gemini 3 Pro sometimes too)
Thanks, but the tweet doesn’t explicitly state that only paid members can use it for free. When other models were first released, was it also only paying users who could use them for free? I’m not very familiar with this, as I’ve always been on the pro plan. https://x.com/cursor_ai/status/1996645841063604711?s=20
The Free plan is not a replacement for the Pro plan as a lot of functionality runs through our servers. This was also the case for all introduction of new models for a while, where they were offered by us or 3rd party providers to paying subscribers as a way to test and learn new models.
I’ve now been waiting for days as it says until Dec 11 its free but I always get the "not enabled in slow pool” error (old 500 requests pricing). So if its free, shouldn’t we be able to use it?