ElevenLabs launches Eleven Realtime with two managed voice-agent models

ElevenLabs introduced Eleven Realtime, a managed voice-agent stack built on Mini and Max models that combines speech recognition, reasoning, turn-taking and voice generation. Max scored 98.8% on Big Bench Audio versus 75.9% for Mini and completed 87.1% of tool-use tasks against 60.3%. Developers cannot select or independently configure the underlying models.

• Max scores 98.8% on Big Bench Audio, Mini 75.9%
• Tool-use tasks: Max 87.1%, Mini 60.3%
• Time to first audio: 570 ms for Max, 620 ms for Mini
• API matches OpenAI Realtime over WebSocket, not WebRTC