Function Calling

An API mode where you declare available functions and the model returns a structured call rather than prose. It is what turns "the model wrote JSON in a sentence" into an actual integration.

Function calling is the mechanism behind tool use. The model still only emits text — the difference is that the text is guaranteed to match the schema you declared, so your code can parse it safely.

The historical significance is worth keeping in view: this feature replaced an era of asking models to emit JSON in prose and parsing it with regexes that broke weekly. Once the provider guarantees the shape, tool use became ordinary software engineering.

• docsFunction calling ↗OpenAI API docs · 2026

  The application-managed tool cycle spelled out: define the function with a JSON Schema, receive the call, execute it yourself, return the output.

• docsFunction calling ↗Google Cloud Vertex AI docs · 2026

  The model selects a declared function and supplies arguments. Useful as a cross-vendor reference showing the pattern is not one provider’s design.

• paperToolformer: language models can teach themselves to use tools ↗Schick et al. (arXiv) · 2023

  Self-supervised learning of when to call an API, which one, with what arguments, and how to use the result — tool use as a learned behaviour rather than a prompt trick.