CLI accepts `--model astra`, shows `astra high`, then reports misleading ChatGPT-account error

What version of Codex CLI is running?

0.155.0

What subscription do you have?

ChatGPT account (exact subscription tier is not exposed by the CLI)

Which model were you using?

astra (user-facing name passed to --model; canonical ID is gpt-6-astra)

What platform is your computer?

Darwin 25.6.0 arm64 arm

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

Ghostty 1.3.1

Codex doctor report

Relevant sanitized fields: Codex 0.155.0; standalone macos-aarch64 installation; ChatGPT authentication; config loaded successfully; installation consistent. Full report omitted because it contains local filesystem paths unrelated to this reproduction.

What issue are you seeing?

The CLI accepts the user-facing model name astra as a raw model identifier and opens normally. The TUI header then displays astra high, which reinforces that the selection is valid.

On the first prompt, Codex prints:

Model metadata for 'astra' not found. Defaulting to fallback metadata; this can degrade performance and cause issues.

It then returns HTTP 400:

The 'astra' model is not supported when using Codex with a ChatGPT account.

The canonical model ID documented for the CLI is gpt-6-astra. The current behavior makes a bad model identifier look like an account-entitlement problem.

What steps can reproduce the bug?

1. Sign in to Codex using a ChatGPT account.
2. Run codex --model astra.
3. Observe that the TUI opens and its header says astra high.
4. Submit any prompt, such as test.
5. Observe the fallback-metadata warning followed by the 400 account-support error.

The exact-error search in this repository returned no existing issue.

What is the expected behavior?

Codex should either:

• resolve the user-facing alias astra to gpt-6-astra; or
• reject the unknown model before opening the session and suggest gpt-6-astra.

The header should not present an unknown, fallback-metadata model as a valid selection, and the error should distinguish an invalid model ID from a subscription-access problem.

Additional information

The attached screenshot is redacted and shows the launch command, the astra high header, the metadata warning, and the 400 response.

Closest related report checked: #45408 uses the canonical gpt-6-astra ID and concerns a client-version/model-catalog rollout mismatch. This report concerns acceptance of the noncanonical astra name and the resulting misleading account-entitlement error.

Start at the CLI's --model handling and follow how the selected name reaches model metadata lookup, TUI header rendering, and the first prompt request. Reproduce with codex --model astra; done means astra is resolved to gpt-6-astra or rejected before the session with a clear invalid-model message, rather than an account-entitlement error.