whichllm
Command-line tool that ranks local LLMs that will run on your hardware
What it is
whichllm detects your GPU, CPU, and RAM, then ranks Hugging Face models that fit, using merged benchmark scores that favor newer model generations over raw parameter count. It estimates VRAM use and tokens per second per quantization, can simulate GPUs you don't own, plan hardware for a given model, and start a chat or print a Python snippet for a pick. People choosing what to run locally, or deciding which GPU to buy, use it to get a short, scriptable answer with JSON or Markdown output.