Rate limit / incomplete responses from AzureOpenAI

[kollerp]

Hi!
We are getting the following errors while interacting with our GPT-5 model:

• Rate limit is exceeded. Try again in X seconds.
• Model did not produce a final response!

How do we understand what is causing this? The only rate limits we can see in the response headers are:

• x-ratelimit-remaining-requests -> We are well below this
• x-ratelimit-remaining-tokens -> This always shows zero (x-ratelimit-limit-tokens is also 0)

Our setup:

• We are using the OpenAI agent SDK (Typescript) passing it an AzureOpenAI client.
• We are streaming responses
• We are running the GPT-5 model