Developer Tools

llama.cpp b9333 adds Apple device ID support for Metal backend

The popular local LLM runner now supports the latest Apple Silicon chips with Metal.

Deep Dive

llama.cpp (ggml-org) released version b9333, adding an Apple device ID for its Metal backend (#23566), co-authored by lvyichen. The release includes builds for macOS Apple Silicon (arm64) with and without KleidiAI, macOS Intel (x64), iOS XCFramework, plus Linux, Android, Windows, and openEuler variants.

Key Points
  • Version b9333 adds a specific Apple device ID for the Metal backend, improving GPU matching on the latest Macs and iOS devices.
  • Release includes builds for macOS Apple Silicon (with/without KleidiAI), Intel Macs, and iOS XCFramework, plus Linux/Windows variants.
  • llama.cpp now has 113,000+ GitHub stars and 18,800+ forks, reflecting its dominance in local LLM inference tools.

Why It Matters

Better Apple Silicon support means faster, more reliable local AI inference for millions of Mac and iOS users.