Using curl to generate images through openAI-compatible mode calls shows 404

I used the following curl command:

curl --location --request POST 'https://generativelanguage.googleapis.com/v1beta/openai/images/generations' \
--header 'Authorization: Bearer $api_key' \
--header 'User-Agent: Apifox/1.0.0 (https://apifox.com)' \
--header 'Content-Type: application/json' \
--header 'Accept: */*' \
--header 'Host: generativelanguage.googleapis.com' \
--header 'Connection: keep-alive' \
--data-raw '{
    "model": "gemini-3-pro-image-preview",
    "prompt": "a portrait of a sheepadoodle wearing a cape",
    "response_format": "b64_json",
    "n": 1,
    "size": "1376x768"
}'

Received the following response:

{
  "error": {
    "code": 404,
    "message": "models/gemini-3-pro-image-preview is not found for API version v1main, or is not supported for predict. Call ListModels to see the list of available models and their supported methods.",
    "status": "NOT_FOUND"
  }
}

I accessed the model list through /v1beta/openai/models and found gemini-3-pro-image-preview included. I also tried other models like gemini-2.5-flash-image-preview, but encountered the same error. Could you please explain what might be causing this issue?

Also, I successfully called it using the URL https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent, but I need to use OpenAI-compatible methods.

Hey! The team is on it, we will get a Nano Banana example added to the docs, might take a few days because of the Holidays but stay tuned.

Hi @Logan_Kilpatrick , any updates on the Nano Banana example or OpenAI-compatible support?

I run into similar issues: Gemini-2.5-flash-image documented in open ai mode but not really supported?

Is the documentation incorrect or my account setting incorrect, really confusing: https://ai.google.dev/gemini-api/docs/openai#generate-image

if the Gemini-2.5-flash-image not really supported in open ai endpoint, why add to documentation in advance?

If this doesn’t supported, why you organization write it in the most important documentaion in advance? It is misguiding the whole industry.

Any updates on this issue ?