Skip to content
llm-speed

Support

llm-speed publishes measured speed and memory numbers for running open-weight LLMs on consumer GPUs and Macs, through llm-speed.com, its API, the llm-speed CLI and the assistant lookups (the llm-speed app in ChatGPT, the Claude connector and the Custom GPT).

Get help

  • Bugs, wrong or missing numbers, feature requests: open an issue on the public repository, github.com/meadow-kun/llm-speed/issues. For a wrong answer from the ChatGPT app, include the question you asked and the llm-speed.com link the answer gave.
  • A model or GPU we haven't measured: request a run at llm-speed.com/requests.
  • Privacy questions: see the privacy page, which lists the privacy contact.

Assistant lookups (ChatGPT app, Claude connector, Custom GPT)

  • They are read-only: they look up llm-speed's published benchmark data and never change anything.
  • There is no sign-in and no account to manage. To stop using the app, remove or disconnect it in your assistant's settings.
  • What the lookups receive and keep is described on the privacy page; the terms cover the data and its CC BY 4.0 license.