Working with OpenAI Chat Completions Models

Working with OpenAI Chat Completions Models | Generated by AI

Home 2025.10

You’ve linked to the Microsoft Learn article “Work with chat completions models” (published on September 1, 2025). Here’s a clear and concise summary of its core points:

Key Highlights

Chat-oriented models The newer chat models differ from earlier GPT-3 models. Instead of “text-in, text-out,” they use a conversation-in, message-out format. Essentially, input is structured like a transcript (roles like system, user, assistant), and the model returns one message to continue the conversation (Microsoft Learn).
Designed for dialogue This structure is optimized for multi-turn back-and-forth conversations. It also works well for single-prompt scenarios, but the conversational format enables smoother interactions (Microsoft Learn).
Don’t treat them like older models These chat models are more verbose and often produce less useful answers if you approach them like GPT-3. It’s recommended to use the new chat-style interface for best results (Microsoft Learn).

Sample code snippet The article includes a basic code sample using Python-style syntax for calling Azure’s chat completions API:

from openai import OpenAI
import os

client = OpenAI(
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),
    base_url="https://YOUR-RESOURCE-NAME.openai.azure.com/openai/v1/"
)

response = client.chat.completions.create(
    model="gpt-4o",  # or deployment_name
    messages=[
        {"role": "system", "content": "Assistant is a large language model trained by OpenAI."},
        {"role": "user", "content": "Who were the founders of Microsoft?"}
    ]
)

print(response.model_dump_json(indent=2))
print(response.choices[0].message.content)

That shows how to properly send messages (with role separation) and retrieve the model’s answer (Microsoft Learn).

Why It Matters

Enhanced conversational context — Chat models better maintain the flow and context across multi-turn interactions.
Improved usability — Using roles (system, user, assistant) allows for clearer instructions, personality setting, or constraints.
Avoid pitfalls — Treating chat models like previous-generation ones can lead to poor response quality; use the chat API interface as intended.

Let me know if you’d like help walking through the code sample, adapting it to your setup, or exploring related guides—happy to dive in!

And as per your preference, I kept this in a simple, readable format without tables.

Back

openai/gpt-5

Donate