# Elevenlabs launches two real‑time speech‑to‑speech models

> Elevenlabs adds a high‑accuracy Max model and a cheaper Mini model for live AI conversations, available via ElevenAgents and a Realtime API.

Oossa · 2026-10-20 · https://oossa.com/en/elevenlabs-launches-two-real-time-speech-to-speech-models

Event date: 2026-10-20

Elevenlabs announced two new speech‑to‑speech models on October 20, 2026. They are called Eleven Realtime v1 Max and v1 Mini. The Max version scored 98.8 % on the Big Bench Audio benchmark and responds in about 0.57 seconds. The Mini version is smaller, costs less, and can detect when a speaker has finished or is only saying “mm‑hmm.” Both models can be accessed through the ElevenAgents platform or an OpenAI‑compatible Realtime API.

## How the models differ

v1 Max is aimed at more demanding tasks such as customer‑support bots that need high accuracy and low latency. v1 Mini is built for natural, everyday conversations where cost is a bigger concern. Pricing is charged per minute of audio plus token usage, with Max priced higher than Mini, according to Elevenlabs’ pricing page.

## The facts

- Elevenlabs released the models on Oct 20, 2026.
- Eleven Realtime v1 Max scored 98.8 % on the Big Bench Audio benchmark.
- v1 Max has a 0.57‑second response time.
- Both models are available via ElevenAgents and an OpenAI‑compatible Realtime API.
- Pricing is per minute of audio plus token costs, with v1 Max costing more than v1 Mini.

## Why it matters

For developers building voice assistants, the Max model offers near‑real‑time replies with very high accuracy, which can improve user experience in support scenarios. The cheaper Mini model lets smaller businesses add live speech interaction without large compute bills. Independent benchmarks beyond Elevenlabs’ own claims have not yet been published.

## Sources & references

1. [Elevenlabs releases two new speech-to-speech models designed for real-time AI conversations](https://the-decoder.com/elevenlabs-releases-two-new-speech-to-speech-models-designed-for-real-time-ai-conversations/) – The Decoder, 2026-10-20

Last updated: 2026-10-20