Magnitude
A local LLM engine runs open-weight models with hardware-specific tuning and connects to coding agents on macOS, Windows and Linux.
Magnitude is a local LLM engine for people who want to run AI agents on their own hardware. It adapts model execution to your machine and connects to coding assistants you already use.
- Hardware tuning: The engine tunes computation for your device and includes optimizations for specific open-weight model families.
- Agent connections: You can use local models with Pi, OpenCode, Hermes and Codex through the desktop app's agent connections.
- Model discovery: The app recommends models and lets you download them within the same interface.
- Concurrent sessions: Agents share cached prompt prefixes to avoid processing the same input repeatedly, and the engine releases their memory when they stop.
Magnitude reports speeds up to twice those of llama.cpp, with separate decode performance gains reported for Metal and CUDA.
The desktop app runs on macOS, Windows and Linux and includes a CLI. It supports Apple Silicon, NVIDIA and AMD GPUs, plus CPU-only operation. Magnitude is free and open source under Apache 2.0, and model inference runs on your own hardware.
Logotyper
