[API] Config.toml = gpt-5.3-codex; silently routes to non-reasoning GPT-5.4 variants

What version of Codex CLI is running?

0.144.1

What subscription do you have?

API

Which model were you using?

gpt-5.3-codex

What platform is your computer?

Ubuntu 24.04.3 LTS

What terminal emulator and version are you using (if applicable)?

_No response_

Codex doctor report

What issue are you seeing?

When Config.toml for Codex CLI and Vscode plugin (both through API key)
set to:

model = "gpt-5.3-codex"
model_reasoning_effort = "xhigh"

All subsequent prompts for the session instead of being routed to gpt-5.3-codex, are routed to a GPT-5.4 variant with reasoning disabled.

----

Error from Codex CLI:

⚠ Model metadata for gpt-5.3-codex not found. Defaulting to fallback metadata; this can degrade performance and cause issues.

Prompt is rerouted to GPT-5.4 variant with reasoning disabled or reduced.

Error from VSCODE Plugin:
No error produced, prompt is silently rerouted to GPT-5.4 variants with reasoning disabled.

What steps can reproduce the bug?

Install latest version of codex cli or use latest version of Vscode Plugin.

Set config.toml to:

model = "gpt-5.3-codex"
model_reasoning_effort = "xhigh"

Restart client.

/model to confirm set model is gpt-5.3-codex

Submit a prompt.

Prompt output responds without reasoning.

What is the expected behavior?

When config.toml is set to:

model = "gpt-5.3-codex"
model_reasoning_effort = "xhigh"

the model should be honored or produce an error consistent with model being deprecated or incorrect.

Additional information

Appears related to https://github.com/openai/codex/issues/32132

_No response_