Artificial intelligence company ElevenLabs has launched two new voice models designed to help businesses build conversational AI assistants and automate customer support. The models can understand spoken questions and respond directly, reducing the need for separate systems to manage voice conversations.
What Are the Two New AI Voice Models?
The company introduced Eleven Realtime v1 Mini and Eleven Realtime v1 Max on October 21, 2026. Both models support more than 90 languages and can use over 17,000 voices from the company’s existing library.
Mini is designed as a lower-cost option for conversational assistants, interactive characters and applications that require natural spoken exchanges. Max is intended for more demanding tasks where AI agents must follow instructions, use software tools and complete a series of steps.
The two models are designed for different requirements, allowing developers to choose between conversational performance and more advanced task execution.
FCRF’s Chief AI Officer Certification Returns on November 14 to Prepare Future AI Leaders
How Do These AI Voice Models Work?
Unlike conventional voice assistants that rely on several separate technologies, the new models combine speech understanding and spoken responses into one system.
Traditionally, developers must connect different components to recognise speech, detect when someone has finished speaking, generate an answer and convert that answer into audio.
ElevenLabs says its speech-to-speech technology handles these functions together, making it easier to develop voice assistants that can respond during conversations.
The approach is intended to reduce the technical complexity involved in building AI-powered customer service systems.
Which AI Model Performs Better?
The two models showed different strengths during performance testing.
In Big Bench Audio, which evaluates spoken reasoning through 1,000 questions, Max answered 98.8% correctly, compared with Mini’s 75.9%.
Max also began speaking slightly faster, averaging 0.57 seconds against Mini’s 0.62 seconds.
For tasks involving software tools, Max achieved an 87.1% completion rate, while Mini recorded 60.3%. These results suggest that Max is better suited to AI assistants that must follow processes and complete tasks involving multiple steps.
However, Mini performed better in tests focused on natural conversation.
Can These Models Handle Real Conversations?
Mini scored 95.8% on Full-Duplex-Bench, compared with Max’s 94.1%. The benchmark measures conversational abilities such as handling interruptions, pauses, turn-taking and brief acknowledgements.
These features are important for voice assistants because human conversations rarely follow a fixed question-and-answer pattern.
In another evaluation, Speech Agent Arena, where people compared two models through live conversations without knowing their identities, Max scored 998 and Mini scored 988.
The results indicate that both models can support interactive voice applications, although their strengths differ depending on the task.
How Can Businesses Use the Technology?
ElevenLabs is targeting developers building customer support assistants and other conversational AI systems.
The Max model is particularly suited to customer service workflows where an AI assistant must access software tools, follow procedures and assess prospective customers through a series of questions.
Mini offers a lower-cost alternative for businesses that primarily need spoken conversations rather than complex task execution.
The company says the models could reduce the effort required to develop voice-based services by combining multiple functions into a single system.
Will Developers Need to Change Their Existing Systems?
ElevenLabs says its developer interface is compatible with OpenAI’s Realtime API, allowing developers to test the new models by changing the connection address and model name in existing code.
However, the interface currently supports only WebSocket connections. Developers using WebRTC, another technology for live communication, would need to change their connection method before using the models.
The420 Insight: What Does This Mean for AI Customer Support?
The launch reflects growing efforts to make AI voice assistants capable of handling both ordinary conversations and more complicated customer service tasks.
The performance differences between Mini and Max also show that faster responses, natural conversations and successful task completion are separate considerations. Businesses adopting such systems should assess which capabilities are most important for their customers rather than relying on a single performance score.
About the author — Ayesha Aayat writes on cybercrime, digital safety, and emerging online threats. Her work focuses on public awareness, legal clarity, and technology-driven risks.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
