What is an LLM API? How Developers Programmatically Access Frontier AI

Overview #

An LLM API is a specialized web API that exposes text, vision, and multimodal inference endpoints for foundation models, accepting conversational message arrays and returning generated completions or token streams over HTTP.

The Distinction Between General AI APIs and LLM APIs #

While 'AI API' encompasses predictive tabular models, computer vision classifiers, and speech-to-text engines, an 'LLM API' specifically refers to generative foundation models. LLM APIs are characterized by conversational state formats, token-based usage meters, temperature and sampling settings, and stream-first responses.

The Industry Standard: /v1/chat/completions #

The OpenAI Chat Completions JSON schema has become the de facto protocol for generative AI applications. Applications submit an array of messages with 'role' ('system', 'user', 'assistant') and 'content'. Unified gateways like API100 preserve this standard schema across all upstream foundation models.

Frequently Asked Questions #

Q: What is the difference between an AI API and an LLM API?

An AI API can refer to any AI/ML service (including OCR, vision, audio). An LLM API specifically refers to generative large language models that take prompts and generate conversational completions.