API（应用程序接口）

The standard interface for applications to access model services, billed by usage.

Model APIs let developers call models over HTTP without running inference themselves. Most vendors expose OpenAI-compatible endpoints with streaming, function calling and context caching—the main delivery channel for closed models.