Error messages

Your organization must be verified to stream this model.

Your organization must be verified to stream this model.

The first sentence is what people search for. The full error continues with a link:

Your organization must be verified to stream this model. Please go to: https://platform.openai.com/settings/organization/general and click on Verify Organization. If you just verified, it can take up to 15 minutes for access to propagate.

From the Python library it arrives as a 400, with the stream field named as the problem (from a bug report, June 2025):

{'error': {'message': 'Your organization must be verified to stream this model. ...', 'type': 'invalid_request_error', 'param': 'stream', 'code': 'unsupported_value'}}

Three versions of the same gate

The wording changes with what's being gated. We found three in GitHub issues and on OpenAI's forum:

• "...to stream this model." The model itself works for you, only streamed replies are locked. A forum thread with over 11,000 views reports exactly that for GPT-5 and GPT-5-mini: non-streaming calls went through, streaming didn't.
• "...to use the model `gpt-5`." The whole model is locked, streaming or not. For image models such as gpt-image-1 the same line comes back as a 403.
• "...to generate reasoning summaries." OpenAI's reasoning guide says "you may need to complete organization verification" before asking for summaries of a model's reasoning.

Paying doesn't get you past it. People in the forum's biggest thread on this had already bought API credit (one had put $50 in before finding out) and hit the gate anyway.

What verifying involves

The link in the error opens your organization's settings, which has a Verify Organization button. It hands you to Persona, an outside identity-check company, which has you take a picture of yourself with your camera (a forum user in that same thread said it's the reason they won't do it). When it succeeds, access can take up to 15 minutes to arrive, per the error itself.

If you'd rather not verify, there are ways round each version. Set stream to false for the first, which costs you the word-by-word display (one user in the big thread said the error stayed with streaming off, so it doesn't work for every account). Drop the summary option from your reasoning settings for the third. For the second, pick a model that isn't gated (the openai-go report found gpt-4.1 working where gpt-5 wasn't).

When you've verified and it still fails

An OpenAI engineer answering that openai-go issue in June 2026 said the error comes "from the API/model access layer rather than the Go SDK", so updating your library won't change anything. His checklist: make sure the API key belongs to the same organization you verified, wait out the propagation window, and if it goes on after that, treat it as an account problem for OpenAI support.

One thing a forum user suggested, which nobody in the thread confirmed and OpenAI hasn't either: create a new API key after verification, on the theory that a key made before verification doesn't pick it up. It's cheap to try. Someone in the Persona thread tried it without luck.

If the Persona step itself breaks (a link with inquiry-id=undefined, or "Session Expired" after a green check), that was a known bug. OpenAI_Support wrote on the forum in November 2025 that they were "working on fixing a bug affecting org verifications now, specifically around regenerating Persona links." People in that thread kept reporting it for two more weeks, until OpenAI_Support said on November 19 that the underlying bug was resolved and asked anyone still unverified to write to support@openai.com. That's still the route if you're stuck now.