ElevenLabs launches Eleven Realtime: a whole voice agent in one speech-to-speech model
Eleven Realtime collapses the usual voice-agent chain — recognition, turn-taking, language model, speech — into two single models, v1 Mini and v1 Max. Each is a complete voice agent, according to ElevenLabs, covering more than 90 languages and the company’s library of more than 17,000 voices.
What happened
- ElevenLabs launched Eleven Realtime on 20 October 2026, with two models: Eleven Realtime v1 Mini for natural, low-cost conversation, and Eleven Realtime v1 Max for complex tasks such as customer support and multi-step workflows.
- Both models support more than 90 languages and any of the 17,000+ voices in the ElevenLabs library, according to the launch post.
- ElevenLabs reports Artificial Analysis results: on Big Bench Audio spoken reasoning, Max answered 98.8% of 1,000 spoken reasoning questions correctly and Mini 75.9%. On tool use, Max completed 87.1% of tasks against 60.3% for Mini.
- Average time to start speaking is 0.57 seconds for Max and 0.62 seconds for Mini, as reported in the launch post — a little over the “half a second” wording in ElevenLabs’ social posts.
- The Realtime API follows OpenAI’s Realtime API specification and connects over WebSocket only; a WebRTC integration would need to move to WebSocket first. Eleven Realtime is also available through the hosted ElevenAgents platform.
- The launch post lists no price for either model. Pricing is per minute plus token costs, with Max priced above Mini, according to ElevenLabs’ pricing page as reported at launch.
What it means
- For creators, Eleven Realtime is a developer product, not a new voiceover button: it matters if you are building a voice agent, a talking character or a phone agent. Narrator-style voiceover work is still ElevenLabs’ text-to-speech offering, covered on the ElevenLabs tool page.
- The Mini/Max split is the decision that matters: Mini for conversation where cost counts, Max where the agent must reason and call tools. The tool-use gap (87.1% vs 60.3%) is the reason ElevenLabs recommends Max for support and multi-step work.
- All benchmark figures here are vendor-reported via Artificial Analysis, as ElevenLabs presented them. No independent leaderboard entry had been published at the time of launch reporting.
What we don’t know yet
- No price was published in the launch post itself. Check ElevenLabs’ pricing page before budgeting a build on either model.
- WebRTC support for the Realtime API was not available at launch, according to the launch post.
How we covered this: facts above come only from the sources listed below. Vendor benchmark figures are the vendor’s own, labelled as such. AiTopMix did not independently test the products in this story.
Sources
Checked on 3 November 2026. If a source page has changed since, the source is the final word.
See the ElevenLabs tool page for the directory’s pricing and free-tier facts, and ElevenLabs vs Murf vs TTSMaker for the voice comparison.