How to stop the model fingerprint from changing

Hi Mason,

Even if you don’t change your model, the system_fingerprint can still change because it reflects the whole environment that runs your model, not just the model itself. The computers, software, and safety systems behind the scenes are sometimes updated or moved around to keep things fast, safe, and reliable. These small background changes—like better hardware, new safety checks, or updated tools—can make the system’s “fingerprint” look different, even though your model and its settings stay exactly the same.

You can minimize changes by:

Make sure you have selected Opt out of automatic model version upgrades as shown below: Go to Azure AI Foundry, select Models + endpoints, and select your model, then click on Edit, click on Customize, and select Opt out of automatic model version upgrades as shown below:

Select the Opt out of automatic model version upgrades:

Below is the python code to check the model_fingerprint:

from openai import AzureOpenAI
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

endpoint = "<MODEL-ENDPOINT>"
deployment = "<DEPLOYMENT-NAME>"
api_version = "<API-VERSION>"

token_provider = get_bearer_token_provider(
    DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)

client = AzureOpenAI(
    api_version=api_version,
    azure_endpoint=endpoint,
    azure_ad_token_provider=token_provider,
)

response = client.chat.completions.create(
    model=deployment,
    messages=[
        {"role": "user", "content": "test"},
    ],
    seed=42,
    max_completion_tokens=5,
)

print(response.system_fingerprint)

Additionally, same responses are not guaranteed even with same fingerprint and seed_parameter. Here is the supported documentation:
https://free.blessedness.top/en-us/azure/ai-foundry/openai/how-to/reproducible-output?tabs=pyton#output-1

Feel free to accept this as an answer.

Thankyou for reaching out to the Microsoft QNA Portal.