The llama.cpp project released version b9927, which includes a significant update to the command-line interface by moving it to an HTTP-based implementation. This change introduces support for router mode within the CLI.

  • The CLI now utilizes an HTTP-based backend instead of previous methods.
  • Router mode is supported in the new CLI implementation.
  • Model aliases are now displayed in the interface.
  • Pre-built binaries are available for macOS (Apple Silicon and Intel), Linux, Android, Windows, and openEuler.
  • GPU acceleration options include CUDA 12/13, Vulkan, ROCm, OpenCL, SYCL, and OpenVINO across supported platforms.

This release provides updated binaries for various operating systems and hardware configurations, alongside the architectural shift in the CLI tooling.