llmfit

Best for: anyone choosing a local model who wants to know what will actually run on their hardware before downloading forty gigabytes.

llmfit, built by Alex Jones, answers a single question with one command: which of hundreds of models and providers fit on your machine. Written in Rust, it covers GGUF, MLX and Unsloth formats and presents results as a fit table in a terminal UI.

What makes it useful: it now reports measured numbers, not just estimates. You can download a model, serve it, benchmark real tokens per second on your own hardware, and contribute the result back as a pull request from inside the TUI — no GitHub CLI or third-party account required. Merged submissions ship in the next release, so anyone on identical hardware gets verified figures instead of a guess.

Where it falls short: coverage of the benchmark database depends on community contributions, so unusual hardware will still fall back to estimates. It tells you what fits and how fast it runs, not whether a model is any good at your task — quality evaluation remains your job.

Details: Rust, MIT licensed. Terminal UI.