whichllm
A command-line tool that detects a machine's GPU, CPU, and RAM, then ranks which local language models actually run well on it using live benchmark data.
whichllm downloads current benchmark data for models on Hugging Face and ranks the ones that fit a detected or simulated GPU, CPU, and RAM configuration, rather than ranking purely by parameter count or file size. Its default mode is described by the project as ambitious, including partial RAM offload and near-the-limit VRAM fits that still look usable; a safer preset restricts results to models that fit fully in GPU memory with a chosen amount of headroom.
A user can simulate a GPU they do not yet own to evaluate a potential purchase before buying it, and results reflect live Hugging Face data rather than a fixed, static list. Because ranking accounts for a model's total and active parameter counts together, a smaller or newer model can outrank a larger one that technically fits.
It is written in Python, distributed under the MIT license, and installable with uv, Homebrew, or pip.