Apple’s Foundation Models CLI: Running Apple Intelligence From Your Terminal

A complete hands-on guide to the new fm CLI in macOS 27.0 Golden Gate

Image: Apple on-device AI via fm

Apple Intelligence has quietly grown a proper command-line interface. On macOS 27, alongside the usual Siri/Spotlight/Writing Tools surface area, there’s a system binary at /usr/bin/fm - the Foundation Models CLI. It's a direct terminal front-end to the same on-device and Private Cloud Compute (PCC) models that power Apple Intelligence system-wide, and it's clearly built for developers, not end users: structured JSON output, an OpenAI-compatible REST server, transcript persistence, token accounting, all callable from a shell. I spent an afternoon putting every subcommand through its paces on a real macOS 27.0 (build 26A5378n) machine. This post walks through what fm actually does, backed by real output - no marketing copy, no guessing.

First look

% fm --help

USAGE
    % fm <command> [options]

COMMANDS
    available     Check model availability
    chat          Start an interactive chat session
    quota-usage   Check model quota usage
    respond       Generate a response to a prompt
    schema        Generate a structured output generation schema
    serve         Start a Chat Completions API server
    token-count   Count tokens in a…