Last updated 2026-09-10
runlocal (the CLI and runlocal.sh) is an informational hardware-fit tool operated by Raybeam. It estimates whether a given open-weight model will run on your hardware, and at roughly what speed. It does not run, host, or serve any model on your behalf — everything it recommends installing runs entirely on your own machine, through third-party runtimes (Ollama, mlx-lm) that you install yourself.
Every fit and speed verdict is a calculated estimate from published model architecture and hardware specs — not a measurement of your specific machine under your specific conditions. Actual results vary with OS load, other running processes, thermal throttling, and driver/runtime versions. A "comfortable" verdict is not a promise; a "tight" or "won't fit" verdict that you choose to override anyway is a decision runlocal will let you make, but the outcome is yours.
Nothing on this site or in the CLI is financial, investment, or purchasing advice. Hardware-fit and "which rig runs this" pages state facts about compatibility; the decision to buy, rent, or use particular hardware is yours to make.
Installing a model may trigger downloads from Hugging Face, Ollama, or mlx-community, and some hardware pages link to third-party retailers (see affiliate disclosure). Those services have their own terms; we don't control and aren't responsible for them.
runlocal is provided "as is," without warranty of any kind. The CLI's source is MIT-licensed — see the LICENSE. To the extent permitted by law, Raybeam disclaims liability for any damage arising from use of runlocal, including from following a verdict that turns out to be wrong.
These terms may be updated as the product changes; the date at the top of this page always reflects the current version.
Open an issue on GitHub.