No legacy models when on Pro

@emsi, which legacy models are you looking for? We retire models after a while, as we cannot support them indefinitely. For a full list of recommended and alternative models that we support, refer to this documentation.

Hi!

@etraut-openai thank you for your response.
It'd be be great to have access to thees:

<img width="590" height="922" alt="Image" src="https://github.com/user-attachments/assets/7cd80230-5264-4940-8ea0-8600fef6253a" />

The "Pro" models don't make sense for Codex usage. When you use a "Pro" model, it's effectively running a model multiple times and choosing the best result. That's great for difficult research-like queries, but doesn't make sense for code generation.

The gpt-5.1 and gpt-5 models are supported in Codex, and you can combine them with various reasoning efforts. Refer to the documentation I linked above for all of the models supported by Codex.

Currently running this command on Pro subscription forces gpt-5.2 and there is no way to choose gpt-5:

$ ./codex -m gpt-5
╭────────────────────────────────────────────╮
│ >_ OpenAI Codex (v0.79.0)                  │
│                                            │
│ model:     gpt-5.2 high   /model to change │
│ directory: ~/Documents/Chats               │
╰────────────────────────────────────────────╯

  Tip: Start a fresh idea with /new; the previous session stays in history.
 
 
  Select Model and Effort
  Access legacy models by running codex -m <model_name> or in your config.toml
 
  1. gpt-5.2-codex (default)  Latest frontier agentic coding model.
  2. gpt-5.1-codex-max        Codex-optimized flagship for deep and fast reasoning.
  3. gpt-5.1-codex-mini       Optimized for codex. Cheaper, faster, but less capable.
› 4. gpt-5.2 (current)        Latest frontier model with improvements across knowledge, reasoning and coding
 
  Press enter to select reasoning effort, or esc to dismiss.  

The "Pro" models don't make sense for Codex usage. When you use a "Pro" model, it's effectively running a model multiple times and choosing the best result. That's great for difficult research-like queries, but doesn't make sense for code generation. The gpt-5.1 and gpt-5 models are supported in Codex, and you can combine them with various reasoning efforts. Refer to the documentation I linked above for all of the models supported by Codex.

I'm talking about Pro subscription, not "pro" models. I need access to GPT-5 and GPT-5.1.

<img width="814" height="336" alt="Image" src="https://github.com/user-attachments/assets/34964241-bdbf-4aab-9d80-c4a4c6fae336" />

The gpt-5.1 and gpt-5 models are supported in Codex, and you can combine them with various reasoning efforts. Refer to the documentation I linked above for all of the models supported by Codex.

As you can see from my comments before, choosing gpt-5 (by -m gpt-5) does not work as expected and runs codex with gpt-5.2.

I'm talking about Pro subscription, not "pro" models. I need access to GPT-5 and GPT-5.1.

Yep, I understood that. My comment above was referring to the two "Pro" models that you highlighted in your screen shot ("GPT 5.1-Pro" and "GPT 5-Pro").

Yeah, it looks like we recently removed support for gpt-5. I'll update the documentation accordingly. The GPT-5 models is pretty outdated at this point (two generations old). The newer models are much more capable. I'm curious why you'd want to use such an old model?

I'm curious why you'd want to use such an old model?

Mainly because it's more capable (willing to undertake more effort) than newer models.

Yeah, it looks like we recently removed support for gpt-5

Not really. When running on different subscription i can access gpt-5, no issues:

$ codex -m gpt-5
╭────────────────────────────────────────────╮
│ >_ OpenAI Codex (v0.79.0)                  │
│                                            │
│ model:     gpt-5 medium   /model to change │
│ directory: ~                               │
╰────────────────────────────────────────────╯