VS Code extension shows API model gpt-5.3-codex as "Custom" and CLI falls back to missing metadata

What version of the IDE extension are you using?

26.715.61943

What subscription do you have?

Logged in with API key

Which IDE are you using?

VS Code

What platform is your computer?

Linux 5.4.0-216-generic x86_64 x86_64

What issue are you seeing?

I am using the OpenAI Codex VS Code extension with API key authentication.

The model gpt-5.3-codex is available to my API account. It is listed by the Models API (GET /v1/models) and can be successfully invoked from the Codex CLI.

I configured the following in ~/.codex/config.toml:

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

When I start a new Codex session, the model works correctly, but the VS Code extension displays the selected model as "Custom" instead of gpt-5.3-codex.

Additionally, the CLI prints the following warning:

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

The session still succeeds and the prompt is answered correctly, indicating that the model is accepted by the API and is actually being used.

What steps can reproduce the bug?

Log into the Codex extension using an OpenAI API key.
Configure ~/.codex/config.toml:
model = "gpt-5.3-codex"
model_reasoning_effort = "high"
Restart or reload the VS Code extension.
Start a new Codex chat or session.

Observed behavior:

The model selector shows "Custom" instead of gpt-5.3-codex.
The CLI prints:
⚠ Model metadata for gpt-5.3-codex not found. Defaulting to fallback metadata; this can degrade performance and cause issues.
Prompts are still executed successfully using gpt-5.3-codex.

What is the expected behavior?

If a model identifier is valid, supported by the API, and explicitly configured by the user:

The VS Code extension should display the configured model name (for example, gpt-5.3-codex) instead of the generic label "Custom".
The CLI should include metadata for the model (or otherwise avoid the fallback metadata warning if the model is officially supported).
If the model is intentionally supported only through API-key workflows, that behavior should be reflected consistently in the IDE.

Additional information

Environment:

VS Code extension: 26.715.61943
Authentication: API key
IDE: VS Code
Platform: Linux 5.4.0-216-generic x86_64
Codex CLI: 0.145.0

Additional observations:

gpt-5.3-codex is returned by the /v1/models API endpoint.
Running the CLI explicitly with:
codex --model gpt-5.3-codex

starts a session showing:

model: gpt-5.3-codex high

and prompts execute successfully.

This appears to be primarily a metadata/UI issue rather than a model availability issue.