Google Gemini 4 Argon available on Databricks Unity Gateway

Google Gemini 4 Argon is a multimodal reasoning model developed by Google that supports text, image, video, and audio inputs. It can be accessed in Databricks using Unity Gateway or Foundation Model APIs.

Use cases

Imagine you need to process complex unstructured data like video and audio while combining advanced reasoning and function calling. With this feature, you can:

• Analyze multimodal content directly within Databricks.
• Execute complex coding and reasoning tasks.
• Run workloads using pay-per-token endpoints.

How to use

You can interact with the model using its managed endpoint or through Unity Gateway.

import mlflow.openai

# Example call using the model service name
model_name = "system.ai.databricks-gemini-4-argon"

# Send your request to the chat endpoint
response = mlflow.openai.chat(
    model=model_name,
    messages=[{"role": "user", "content": "Analyze this audio file and extract the main points."}]
)
print(response)

What happens in this code:

1. Defines the model service identifier system.ai.databricks-gemini-4-argon.
2. Sends a prompt to the Unity Gateway managed endpoint.
3. Outputs the response generated by the model.

Limitations and requirements

• The model uses a global endpoint and requires cross-geography routing to be enabled.
• Reasoning controls are enabled and cannot be disabled.
• Customers are responsible for ensuring compliance with applicable model terms.
• Supported under pay-per-token mode in supported regions.