ElevenLabs launches faster AI voice models for live phone-style conversations

ElevenLabs has released two new AI voice models built for live, back-and-forth conversation. They are called Eleven Realtime v1 Max and v1 Mini. Both listen to speech and answer in speech. The company says the top model is the fastest and most accurate of its kind on one public leaderboard.

What changed

The v1 Max model scores 98.8 percent on the Big Bench Audio test and responds in 0.57 seconds. ElevenLabs says no model on the Artificial Analysis leaderboard is both faster and more accurate right now. That is the company's claim. The material does not show independent testing.

The v1 Mini is meant for natural conversation at lower cost. It can tell when someone has finished talking and when they are just saying "mm-hmm." That helps an AI avoid cutting in at the wrong moment. V1 Max is aimed at harder jobs, such as customer support.

Who it is for and what it costs

Both models run on the ElevenAgents platform and through an OpenAI-compatible Realtime API, a connection that works like OpenAI's. Pricing is per minute plus token costs, which are usage-based charges. V1 Max costs more than v1 Mini. Exact prices are on the ElevenLabs pricing page.

Businesses thinking about an AI that answers calls are the clear audience. What is not yet known is how these models perform on real customer calls, or how the total cost compares with other options.