Jev latency protocol comparison reads two descriptions of what was timed. A hosted API round trip and a local model forward pass can have the same request text while including different work. Our original question checks stated workload and start/end boundaries. It does not measure a model or calculate which provider is faster.
An API-collected X discussion objected to comparisons that mix hosted access with local execution. That motivated six original protocol descriptions. We do not repeat the source's speed multiplier, reproduce an outside benchmark or claim a current hardware result. The real API measurements displayed here concern the protocol-review calls themselves.
Comparable stated protocol is deliberately narrow. It says that the two supplied descriptions use the same workload and boundary. Hardware, batching, cache behavior, network conditions and answer quality can still differ. The fixed report keeps measurementsVerified false so a wording check cannot become a verified performance claim.