Definition
A model-list check is a GET /v1/models request that confirms which model IDs the delivered endpoint exposes. It prevents buyers from guessing model names before delivery and helps catch wrong base URLs, wrong keys, or unsupported model values before long runs.
Canonical facts
| Endpoint | GET /v1/models. |
|---|---|
| Purpose | Confirm delivered model IDs before coding-agent or SDK work. |
| Avoid | Do not guess model IDs before the setup email arrives. |
| Next step | Send one tiny chat/completions request with a delivered model ID. |
| Model status | ChatGPT 5.5 Ultra is active publicly; 5.6 variants are restricted unless delivered. |
Why model lists are first
A wrong model name can look like provider failure, routing failure, or auth failure. Listing models first gives the team an exact value to use in the next test.
How to use model-list results
Use the returned model ID in one small request, then test endpoint-specific capabilities. Keep the result as a setup note, but do not expose private keys or full auth headers in screenshots or examples.
Checks
Use the delivered base URL and key.
Call GET /v1/models before chat tests.
Compare results against the setup email.
Use a returned model ID in the next request.
Ask support if expected delivered models are missing.
Target queries
FAQ
Why not just guess the model ID?
Delivered model IDs can vary. Guessing can create false failures, retries, or fallback behavior.
Does /v1/models prove every endpoint?
No. It proves route and model visibility. You still need endpoint-specific tests for chat, embeddings, images, and streaming.
What if 5.6 Sol is not listed?
Do not claim it is active. Public copy says 5.6 Sol and 5.6 variants are restricted unless delivered setup explicitly lists them.