Magnitude Runs Open Models 2x Faster Than Ollama

Magnitude, a free open-source desktop app released by Tom Greenwald, automates GPU thread and memory configuration to run local AI models up to 2.19x faster than Ollama on Mac M4 Pro and 2.17x faster on DGX Spark. It supports Apple Silicon, Nvidia, AMD, and CPUs, uses a quantized KV cache to halve memory usage, and plans future updates including attention pruning and expert streaming.