tokens per second

yes / slow / no verdicts with estimated tok/s for every local LLM on your hardware, before you download. Then verify, serve and chat, locally.