Error messages

The model `gpt-5.5` does not exist or you do not have access to it.

The model `gpt-5.5` does not exist or you do not have access to it.

The model name in backticks is whatever the request asked for. Here it's one from a September 2026 report:

The model `gpt-5.5` does not exist or you do not have access to it.

It comes back as HTTP 404 with this body:

{
  "error": {
    "message": "The model `gpt-5.5` does not exist or you do not have access to it.",
    "type": "invalid_request_error",
    "param": null,
    "code": "model_not_found"
  }
}

Why the message is vague

The API gives the same answer for "no such model" and "not for you", so the text alone won't say which. There are four likely causes, and it's quicker to rule them out in order than to guess.

First, check what your key can see

OpenAI's List models endpoint "lists the currently available models" for the key you call it with:

curl https://api.openai.com/v1/models -H "Authorization: Bearer $OPENAI_API_KEY"

Search the output for your model's exact ID. If it's there, the problem is in the request (a stray space, a capital letter, a display name like "GPT-5.5" where the ID is lowercase). The reply also carries a shutdown_date for each model, which is worth a look while you're there.

A model that's been retired

Once a model hits its shutdown date, OpenAI's deprecations page says it "will no longer be accessible," and requests for it get this 404. The next batch is close: on September 28, 2026, gpt-3.5-turbo-instruct, babbage-002, davinci-002 and gpt-3.5-turbo-1106 shut down, with gpt-5.6-terra as the listed replacement. Dated GPT-5 and o3 snapshots follow on December 11, 2026.

Preview models go faster. The same page says anything with "preview" in the name "may be retired with much shorter notice, such as 2 weeks", and names computer-use-preview as an example. It was shut down on July 23, 2026, so every request for it now gets this 404. (It also turns up in an August 2025 issue on OpenAI's own sample app, where people following the README got the same line nearly a year before the shutdown, with no explanation in the thread.)

A model your account can't use yet

Some models are held back from newer or smaller accounts. In 2024 the same error for gpt-4 and o1-mini turned up in GitHub trackers, and the maintainer of one code-review tool told a user "it's the OpenAI limit that you don't have access to o1-mini for your API key." OpenAI moves accounts up usage tiers automatically as they spend, starting from $5 paid for Tier 1. Models behind organization verification usually give a different, clearer error.

The wrong key or the wrong server

If you belong to more than one OpenAI organization, the key in your environment may come from one that doesn't have the model (an OpenAI engineer gave the same first check for a related access error). And the wording isn't unique to OpenAI: Groq copies OpenAI's API format and returns the same line (one September 2026 report got it for llama-3.3-70b-versatile after Groq retired that model), so a tool pointed at a different base URL will complain about an OpenAI model name that server has never heard of.