Bindwidth
A browser-based calculator that finds which hardware limit actually constrains an on-premises LLM deployment, then prices and compares sizing options against it.
Bindwidth is a frontend-only calculator for sizing a private, on-premises large language model deployment. It evaluates three limits that can constrain a single resident text model, memory for the key-value cache, combined prefill and decode serving capacity, and the runtime session ceiling, identifies which one actually binds for a given team and model, and prices and compares eligible hardware and hosting alternatives against that constraint.
A guided setup walks through team shape and model choice to a binding-constraint result in a few seconds, and every hardware and model profile carries a confidence level distinguishing measured figures from estimated ones, which is included in exported reports. Results can be exported as a Markdown decision record or as a JSON scenario file that can be reloaded later, and a request-quote flow copies a summary of the sized configuration into a form.
It runs entirely in the browser with no build step, server, database, account, or telemetry, and is MIT licensed.