How do I stream chat completions with OpenAI’s Python API?

[Istituto-freudinttheprodev]

I'm using the official openai Python package to call the Chat API (gpt-3.5 or gpt-4), and I'd like to stream the response instead of waiting for the full reply.

I tried this:

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello"}]
)

But it waits until everything is returned.

How can I make it stream the tokens one by one as they’re generated?

[MatteoMgr2008]

Great question!

To stream chat completions using the openai Python package, you need to set stream=True and then iterate over the events.

Here’s how you can do it:

import openai

openai.api_key = "your-api-key"

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True  # ✅ this enables streaming
)

for chunk in response:
    if "choices" in chunk:
        content = chunk["choices"][0]["delta"].get("content", "")
        print(content, end="", flush=True)

This will print the generated message token-by-token in real time.

Let me know if that works — and feel free to mark this as the answer if it helps! ✅

[Istituto-freudinttheprodev]

Thank you so much!

[AhmedGMurtaza]

here is a detailed explanation : https://cookbook.openai.com/examples/how_to_stream_completions

[MatteoMgr2008]

@AhmedGMurtaza Thanks for the link to the article, I didn't think OpenAI had released an official paper on this specific topic!
In any case I will gladly read it to find out more