ElevenLabs releases Eleven Realtime speech-to-speech models for live AI conversations

Eleven Realtime arrives in two variants

ElevenLabs has released two speech-to-speech models under the name Eleven Realtime, designed for real-time AI conversations. The company says the larger v1 Max variant scores 98.8 percent on the Big Bench Audio benchmark and has a response time of 0.57 seconds. ElevenLabs says that, at the time of publication, no model on the Artificial Analysis leaderboard is both faster and more accurate.

On the Artificial Analysis Speech to Speech leaderboard, Eleven Realtime v1 Max is placed near the top left corner, a position that reflects high accuracy combined with low latency. The company published the comparison as part of its launch material.

Mini focuses on turn taking and cost

The smaller v1 Mini is built for natural conversations at lower cost. ElevenLabs says it can tell when someone has finished talking and can also distinguish that state from a brief mm-hmm, which could reduce false interruptions in spoken exchanges. The Max variant is aimed at more complex tasks such as customer support.

Availability and pricing

Both models are available through the ElevenAgents platform and through an OpenAI-compatible Realtime API. Pricing is charged per minute plus token costs, with v1 Max costing more than v1 Mini. ElevenLabs says full details are on its pricing page.

Sources: The Decoder