A focused utility for the self-host question.

RunWhere helps practitioners decide whether an open-weight model belongs on their own cloud GPUs or behind a hosted API. It is directional by design: the answer is a regime, not a spreadsheet.

It is the third worked example in the Precomputed AI trilogy, following RightModel for ruleset compilation and CloudEstimate for scheduled generation.

The project has no accounts, no billing, and no persistence beyond stable shareable URLs.