Voice agents: ElevenLabs launches faster speech-to-speech models
In brief
On 20 October 2026, ElevenLabs announced Eleven Realtime v1, a pair of speech-to-speech models, Mini and Max, available the same day. According to the company, they replace a separate speech-to-text, language model and text-to-speech chain with one native voice-agent model. They support up to 90 languages and ElevenLabs' 17,000 voices, and the Realtime API is compatible with OpenAI's.
What it changes for an SME
This launch does not change your compliance duties. If you already trial voice agents, you can compare Mini for simple talks and Max for multi-step reasoning or tools and, according to ElevenLabs, switch from OpenAI's Realtime API by changing the endpoint and model name. Cost is still unclear: ElevenAgents bills per minute, but token prices for the API had not been published. A reported delay of 470 to 500 milliseconds may shorten the wait on a phone line, without proving it will fit your real calls.
What you can do
- Check whether your current voice flows still work after changing only the API endpoint and model name, as ElevenLabs describes.
- Ask for the per-minute cost estimate in ElevenAgents and wait for token prices before you set a budget.
- Use Mini for simple dialogues and Max if calls need tools or multi-step reasoning, according to ElevenLabs.
Let's talk about the time you could save.
In 20 minutes, we find where AI could save you time or money. Free, no commitment.