›_ Codex Issues
Codex Issues / #34967

Regression in 26.715.72359: Automatic approval broken for all GPT-5 models (Microsoft Store)

Open 💬 3 comments Opened Jul 23, 2026 by jojoaurelien
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of the Codex App are you using (From “About Codex” dialog)?

26.715.72359

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

After updating Codex Desktop to version 26.715.72359, Automatic approval stopped working completely.

Every request requiring automatic approval fails immediately with:

This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite

I tested multiple models:

  • GPT-5.4
  • GPT-5.5
  • GPT-5.6 Sol
  • GPT-5.6 Terra
  • GPT-5.6 Luna

All fail with the same error.

Switching to "Ask for approval" works normally.

This issue started immediately after updating to 26.715.72359. The exact same workflows worked correctly before the update.

This is not related to my project or my code. It happens in every project requiring network approval.

What steps can reproduce the bug?

  1. Install Codex Desktop 26.715.72359 from the Microsoft Store.
  2. Open any trusted project.
  3. Enable "Automatic approval".
  4. Select GPT-5.4, GPT-5.5 or any GPT-5.6 model.
  5. Ask Codex to perform an action requiring network access (for example an HTTP request, API call, or package installation).

Result:

Automatic approval fails immediately with:

This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite

Additional tests performed:

  • Restarted Codex
  • Switched between all GPT-5 models
  • Edited ~/.codex/models_cache.json (use_responses_lite=false)
  • Tried making the cache read-only
  • Created a custom model catalog

None of these solved the issue because the cache is regenerated automatically.

What is the expected behavior?

Automatic approval should work exactly as it did before version 26.715.72359.

When Codex needs network access, the automatic approval review should succeed and allow the task to continue without requiring manual intervention.

No X-OpenAI-Internal-Codex-Responses-Lite compatibility error should occur.

Additional information

Environment

  • Codex Desktop
  • Version: 26.715.72359
  • Installed from Microsoft Store
  • Windows 11
  • ChatGPT Plus

Impact

This is a critical regression.

I use Codex Desktop professionally with autonomous coding workflows.

Many of my routines perform dozens (sometimes hundreds) of network operations.

Automatic approval is an essential feature.

Because every network request now requires manual approval, these workflows are no longer usable.

The application worked correctly before updating to version 26.715.72359.

Several GitHub issues mention X-OpenAI-Internal-Codex-Responses-Lite, but in my case every GPT-5 model is affected (5.4, 5.5, 5.6 Sol, Terra and Luna).

View original on GitHub ↗

3 Comments

github-actions[bot] contributor · 1 month ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #34936
  • #34872

Powered by Codex Action

noah-foltz · 1 month ago

Seconded, but on codex-cli 0.145.0 on Linux Mint

marcelodiaz558 · 1 month ago

I am also facing this issue on the following two platforms:

  1. codex-cli 0.145.0 on Linux Mint
  2. codex-cli 0.145.0 on Ubuntu 20.04.5 LTS

This is the issue I see on Codex CLI when using auto-approve.

Automatic approval review denied (risk: high, authorization: unknown): Automatic approval review failed: {
    "type": "error",
    "error": {
      "type": "invalid_request_error",
      "code": "unsupported_value",
      "message": "This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite.",
      "param": "model"
    },
    "status": 400
  }

Another weird thing I note is that when I use /model I see the following:

Select Model
  Pick a quick auto mode or browse all models.
 
  1. codex-auto-review     Balanced agentic coding model for everyday work.
› 2. All models (current)  Choose a specific model and reasoning level (current: gpt-5.6-sol)

I guess codex-auto-review should not be an option.

Bear in mind that I tried to close codex CLI and start it again and the auto-review seems to be working. I'll write another comment here if it happens again.