# llmfit

A hardware-detection tool that inspects a machine's CPU, RAM, GPU and VRAM to recommend which open-source language models, and which quantization, will actually run well on it.

llmfit detects CPU cores, system RAM, discrete and integrated GPUs, VRAM, and unified-memory architectures such as NVIDIA CUDA, Apple Silicon, AMD ROCm and Intel OneAPI, then scores candidate models across quality, speed, fit and context length using their parameter counts, context lengths and quantization formats such as GGUF, AWQ, GPTQ and EXL2. It supports multi-GPU setups and mixture-of-experts architectures, and can suggest a runtime provider such as Ollama, llama.cpp, MLX, Docker Model Runner or LM Studio.

It offers an interactive terminal UI by default, a classic CLI mode, a web dashboard, and a REST API for integration into other tooling. Users can benchmark a downloaded model on their own hardware and contribute the measured tokens-per-second numbers back to the project, replacing its estimates with real data over time.

It is written in Rust, distributed for macOS, Linux and Windows through Scoop, Homebrew and MacPorts, and licensed under the MIT license.

Website: https://github.com/AlexsJones/llmfit