●Change8

v0.30.0-rc21

Breaking Changes
📦 ollamaView on GitHub →
⚠ 1 breaking✨ 3 features

Summary

This pre-release version overhauls the architecture to use llama.cpp directly, enabling GGUF support and leveraging MLX for Apple Silicon acceleration. Users are encouraged to provide feedback on performance and stability.

⚠️ Breaking Changes

  • The underlying architecture has changed from using GGML to directly supporting llama.cpp, which may affect custom integrations or tooling relying on the previous internal structure.

Migration Steps

  1. If you rely on the previous GGML architecture, be aware that it has been replaced by direct llama.cpp support.
  2. To install this specific pre-release version (0.30.0-rc21) on Mac/Linux, use: curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION=0.30.0-rc21 sh
  3. To install this specific pre-release version (0.30.0-rc21) on Windows, use: $env:OLLAMA_VERSION="0.30.0-rc21"; irm https://ollama.com/install.ps1 | iex

✨ New Features

  • Direct support for llama.cpp architecture.
  • Compatibility with the GGUF file format is now enabled.
  • MLX is utilized to accelerate model inference on Apple Silicon devices.