ElevenLabs announced STS Mini and Max through its API and ElevenAgents. The speech-to-speech models support more than 90 languages and varied voices, targeting real-time turn-taking, interruptions and tool calls. They offer an OpenAI Realtime-compatible WebSocket connection but do not support WebRTC at launch. Published latency and accuracy figures reflect the company’s evaluation conditions; practical performance needs testing with actual networks, languages and tool execution.