Speko is a platform for building and deploying voice agents, with a pinned model stack for each agent. You can run separate listen, think, and speak stages in Cascaded mode, or a single realtime pass. This integration makes Smallest AI's Pulse the speech-to-text stage, so your agent hears callers accurately and, just as important, knows when they have stopped talking. End-of-turn latency is the wait between a caller finishing and the agent replying, and it is what makes a live agent feel responsive or sluggish. Speko's own public benchmarks rank Pulse first on end-of-turn finalization among the streaming models it tests, which is why it is a natural default for the STT stage in a production Speko agent.