ElevenLabsNotable

Eleven Realtime v1 Max

Released Oct 19, 2026

ElevenLabs' reasoning speech-to-speech model for voice agents. It bundles speech recognition, turn-taking, an LLM and voice in one model choice, and reasons before it answers.

Price per minute: $0.080/min
Inputs: audio, text
Tool calling: Yes
Languages: 90
Voices: —
Big Bench Audio: 98.8%
Access: Generally available
Licence: Proprietary

Strengths & weaknesses

Strengths

Big Bench Audio 98.8% and 0.57s time to first audio (AA)
Completed 87.1% of tool-use tasks
Realtime API compatible with OpenAI's over WebSocket
90+ languages and the ElevenLabs voice library (17,000+ voices)

Weaknesses

Token charges ($1.80 in / $9.00 out per 1M) on top of $0.08/min connected audio
Full-Duplex-Bench 94.1%, slightly below the cheaper Mini

Notable: top 10 on both AA boards (#7 S2S Index 77.3; #4 Big Bench Audio 98.8%). releaseDate is the changelog launch (gradual rollout from 2026-10-19, all workspaces by 2026-11-02); the blog post is dated 2026-10-20. Benchmarks from ElevenLabs' blog, measured by AA: Speech Agent Arena 998. The dashboard default for Realtime voice mode since 2026-10-26. perMinuteUsd is the connected-audio rate only; token charges apply on top.

Links

API model strings

elevenlabseleven_realtime_v1_max