Eleven Realtime Eleven Realtime is a new class of end-to-end speech-to-speech models for voice agents. It starts speaking in about half a second while handling spoken reasoning accurately, combining recognition, turn-taking, an LLM, and voice output in one system. v1 Max scores 98.8% on Big Bench Audio with 0.57s latency and supports complex tool-using tasks; v1 Mini is optimized for natural, low-cost conversation including interruptions. Available today in ElevenAgents and via an OpenAI-compatible Realtime API supporting 90+ languages.
Visit productFeatured media
Loading media…
View on XCosigns Received
“MulmoChat works with Eleven Realtime beautifully! MulmoChat is an open source multi-modal chat application, which supports "GUI chat protocol". https://github.com/receptron/MulmoChat”
“Launch voice agent experiences with ElevenLabs easier than ever. Literally just changing base url in your realtime sdk's init! Great ship by the team with a lot of improvements under the hood.”
“i recently spent HOURS on hold with my auto insurance company trying to reach someone who could help me imagine a world where you can call your insurance, your doctor’s office, or a government agency and immediately have a useful conversation with an agent who understands what you need this is the kind of AI progress i’m excited about”
“Voice agents that feel like real conversation need two things: speed and accuracy. Eleven Realtime, our new speech-to-speech model, delivers both, with responses in ~0.5s. If you are building something with voice, try it out today!”
“New ElevenLabs model. Our fastest yet for voice agents - Eleven Realtime listens, thinks and replies in under half a second.”