Short answer
As of November 10, 2026, don’t build a new integration around gpt-5.2-codex: OpenAI’s API model page marks it “Shut down.” The page identifies gpt-5.2-codex as the model ID, but that is not a reason to expect a request using it to work now. (developers.openai.com)
That changes the answer to “How do I make my first API request?”: there is no current, supported first request to give for this shut-down model. Choose a model that OpenAI currently lists as available, then use that model’s own API documentation for the endpoint, parameters, and request example. Don’t substitute a third-party gateway’s model name or adapt a sample for a different model as if it were verified for GPT-5.2-Codex.
What the official API page says
OpenAI’s model page lists the API identifier as gpt-5.2-codex and explicitly labels the model “Shut down.” The same page lists a 400,000-token context window, a 128,000-token maximum output, image input, and support for streaming, function calling, and structured outputs. Those are specifications displayed on the page—not a promise that the shut-down model can still be called. (developers.openai.com)
The page also displays rates of $1.75 per million input tokens, $0.175 per million cached input tokens, and $14 per million output tokens. Because the model is marked shut down, don’t use those displayed figures as a current price quote or production budget. Confirm pricing for the active model you select. (developers.openai.com)
Why older API guides can mislead
OpenAI’s launch announcement, dated December 18, 2025, said the company was working toward enabling API access in the coming weeks. That was a statement about the rollout at the time; it does not establish that the model is available now. The current model page’s shut-down label is the more relevant signal for an integration decision. (openai.com)
Third-party pages may still list the model or show a price. For example, a gateway guide lists gpt-5.2-codex but demonstrates a different model, gpt-5-codex. That example does not verify a GPT-5.2-Codex request. Gateway model names, credentials, and endpoints also apply to that gateway’s route—not automatically to OpenAI’s API.
What to do before integrating
- Pick an active model. Check OpenAI’s current API model documentation rather than relying on a search result, old launch announcement, or gateway list.
- Use the exact ID and endpoint documented for that model. Don’t assume changing only
modelin an old request is enough; supported inputs, tools, parameters, and API behavior can vary. - Run a minimal test request with the selected model before wiring it into an application. Confirm that the request succeeds and that your code reads the returned output correctly.
- Recheck your production assumptions. Verify the chosen model’s current prices, limits, and needed capabilities in its documentation before estimating cost or deploying.
Bottom line: gpt-5.2-codex is the identifier shown on OpenAI’s model page, but that page currently marks the model as shut down. Treat it as a historical model reference, not a usable API integration target.
Sources
GPT-5.2-Codex Model | OpenAI API
How to Access GPT-5 and GPT-5.2 via API - Complete Developer Guide