llmfit is a terminal tool that helps you find which large language models will run well on your specific hardware.
The tool solves the problem of model selection by detecting your system's RAM, CPU, and GPU capabilities, then scoring available models across quality, speed, fit, and context dimensions to identify which ones will actually perform acceptably on your machine. It supports hundreds of models and multiple providers including Ollama, llama.cpp, MLX, Docker Model Runner, and LM Studio. The interactive terminal interface displays this information in a searchable format, while a classic CLI mode supports automation. The tool handles multi-GPU setups and models with mixture-of-experts architectures, and can dynamically select quantization levels based on your hardware constraints.
You should adopt this tool if you run local language models and want to avoid the trial-and-error of downloading models that turn out to be too large or slow for your system. It suits anyone managing their own inference infrastructure, from individual developers experimenting with models to teams deploying across heterogeneous hardware. The project includes a benchmarking feature that lets you measure actual tokens-per-second performance on your machine and contribute those measurements back, building a shared database of real-world performance data so others with identical hardware can see verified numbers before running their own benchmarks.
Almost all open issues are raised by outside users rather than the core team, indicating a substantial base of adopters reporting real-world use. Maintainers typically respond to new issues and pull requests within a day. Work in the issue tracker is dominated by bug reports, enhancement requests, and feature requests.
