ElevenLabs Just Dropped Realtime AI Voice: Here's What Changes

According to the-decoder.com, ElevenLabs has expanded its voice technology lineup with two new speech-to-speech models named Eleven Realtime. The launch targets the growing demand for natural, fluid voice interactions in automated systems and customer support channels.

How the Eleven Realtime models work

The lineup consists of two variants designed for different performance needs. The v1 Max variant achieves a 98.8 percent score on the Big Bench Audio benchmark alongside a 0.57-second response time, positioning it at the top of the Artificial Analysis leaderboard for combined speed and accuracy. Meanwhile, the smaller v1 Mini focuses on lower operational costs and can distinguish between conversational fillers like "mm-hmm" and actual completion markers.

How to test and integrate Eleven Realtime

Developers and teams can access the new models directly through the ElevenAgents platform or via an OpenAI-compatible Realtime API. Pricing follows a consumption-based structure charged per minute plus token costs, with the higher-accuracy v1 Max carrying a steeper price tag than the lightweight v1 Mini.

Is it worth it for your workflow?

For support teams and workflow automation builders, sub-second latency eliminates the awkward pauses that previously made phone bots frustrating to use. If your daily operations involve voice-based customer interactions, testing the v1 Mini for high-volume routing or the v1 Max for complex problem-solving offers a practical way to upgrade automated phone trees without sacrificing conversational quality.