OpenAI-Compatible API

The platform's model inference API is compatible with the OpenAI API. This means:

• Almost zero changes to existing code: just replace base_url and api_key to switch an application that calls OpenAI over to a platform endpoint.
• Use the official OpenAI SDKs directly: existing toolchains in the Python and JavaScript ecosystems work as-is; no proprietary SDK to install.
• Public and private endpoints are called the same way: both follow the same API conventions; the only differences are the Base URL and model name.

Get connection details

Public endpoints: in Hub > Models Hub, click a model card and choose API Endpoint. The panel provides Python / JavaScript / cURL code samples and the API path.

Private endpoints: in Gen AI Studio > My Endpoints, click an endpoint. The Private API Information section under Details includes the Base URL, model name, endpoint type, and API path; click Show Code for a complete sample.

Example calls

Using a chat model as an example (use the Base URL and model name shown in the console panel):

from openai import OpenAI

client = OpenAI(
    base_url="<API path shown in the console>",
    api_key="your-api-key-here",  # replace with your API key
)

response = client.chat.completions.create(
    model="<model name shown in the console>",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is a GPU?"},
    ],
    temperature=0.7,
)
print(response.choices[0].message.content)

curl <API path shown in the console> \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your-api-key-here" \
  -d '{
    "model": "<model name shown in the console>",
    "messages": [{"role": "user", "content": "What is a GPU?"}]
  }'

tip

Parameters you tune in the Playground (such as Temperature) are reflected in the sample generated by View code — tune until you are happy, then copy. Easiest path.

Authentication and credits

• All requests are authenticated with an API key; keys can have a Budget Limit, and requests are rejected once the limit is reached.
• Public endpoints are billed by token usage and subject to rate limits; every call record can be found in the usage reports.