Option to change the model within the Plan Mode approval selector

yes this is a real enhancement would love to see this implemented.

What

This patch adds support for overriding the model used specifically in Plan mode via a new config field:

plan_mode_model = "gpt-5.1-codex-mini"

Why

Currently, Plan mode always uses the same model as execution. In practice, it can be useful to:
use a cheaper/faster model for planning
keep a stronger model for execution
This separation improves both cost and responsiveness without affecting execution quality.

Summary of changes

Add plan_mode_model to:
config TOML
profile config
protocol types
JSON schema
Apply override in TUI when entering Plan mode
Preserve correct fallback behavior:
Plan mode falls back to model if unset
Switching back restores the original executor model
Add tests covering:
config deserialization
precedence (profile > top-level)
TUI behavior
snapshot updates
Update docs

Example

model = "gpt-5.4"
plan_mode_model = "gpt-5.4-mini"
Plan mode will use gpt-5.4-mini, while execution still uses gpt-5.4.

Patch

Attached as a .patch file (generated via git format-patch origin/main).
Happy to split or adjust this if needed.
0001-codex-Add-a-Plan-mode-model-override.patch

@Playitcooool Not sure I understand why you sent this patch, your comment talks about using a different model for planning and executing, which is already possible. I am talking about changing the executor model within plan mode, before execution, which your patch does nothing about.

@Playitcooool Not sure I understand why you sent this patch, your comment talks about using a different model for planning and executing, which is already possible. I am talking about changing the executor model within plan mode, before execution, which your patch does nothing about.

Literally, we are talking about the same thing.I came up with patch because I met the same problem when I click "yes, implement the plan" codex directly switches to default model leaving me no chane to change model.After setting an explicit plan mode model, we can achieve codex switching to the executor model we want.

Moreover, I cannot use different models for plan and execution.The situation is that every time you use "/model" for changing a model for planning, codex modifies the config.toml, and after that your default model is changed.So when you start to implement the plan, codex switches to the modified default model which is not I want.

In the update-to-date version, I can only configure different reasoning efforts for plan and execution, but cannot configure different models for them.

@Playitcooool I might be lost now, but my problem specifically manifests as the plan mode being a better thinking depth than the default model, and wanting to bring the executor/default model back up to par. As for your point about it changing the default, when you use /model to change model for planning, it asks you if you want to change it for plan mode only or for both. My issue is that once you hit the selector at the end of a plan mode session you can’t perform any changes to the executor model.

@Playitcooool I might be lost now, but my problem specifically manifests as the plan mode being a better thinking depth than the default model, and wanting to bring the executor/default model back up to par. As for your point about it changing the default, when you use /model to change model for planning, it asks you if you want to change it for plan mode only or for both. My issue is that once you hit the selector at the end of a plan mode session you can’t perform any changes to the executor model.

I am lost, but we share the same issue "once you hit the selector at the end of a plan mode session you can’t perform any changes to the executor model". So once this issue is solved, mine is solved lol.