llmfit

A local LLM selection tool that checks your hardware, ranks models by fit and estimated speed, and benchmarks them through local runtimes.

llmfit helps people choosing models for their own computers compare what will fit in memory and how fast it might run. It checks your hardware before recommending models, so you can assess candidates before downloading them.

Hardware-aware recommendations: Compare models by memory fit, estimated speed, quality, and context length, with support for multiple GPUs and mixture-of-experts models.
Quantization choices: Assess memory and performance tradeoffs across GGUF, AWQ, GPTQ, and EXL2 formats.
Local benchmarks: Measure generation speed and time to first token through Ollama, llama.cpp, MLX, Docker Model Runner, or LM Studio; saved measurements replace estimates in your results.
Terminal, browser, and API access: Browse recommendations in an interactive terminal or self-hosted web dashboard, or use JSON output and an HTTP API in scripts and other applications.

The open-source project uses the MIT license and runs on macOS, Linux, and Windows, with Docker deployment available. Hardware detection covers NVIDIA CUDA, Apple Silicon unified memory, AMD ROCm, and Intel OneAPI.

Performance figures are estimates unless backed by measurements. Benchmark runs are saved locally first, and sharing results is optional. The program contacts external services only when you request a feature that uses them, such as model downloads, runtime provider queries, or the community leaderboard.