Elevenlabs launches low-latency speech-to-speech models

Elevenlabs released two new speech-to-speech models called Eleven Realtime on 20 October 2026. According to The-decoder, the company launched these models to facilitate real-time AI conversations. The v1 Max model achieved a 0.57-second response time and scored 98.8 percent on the Big Bench Audio benchmark. The smaller v1 Mini version focuses on lower costs for natural dialogue. Both models are now accessible via the ElevenAgents platform and an OpenAI-compatible Realtime API.

Why it matters

Swiss businesses can now deploy voice AI that reacts in under one second. This speed removes the awkward pauses that often break user trust in automated systems. You can choose between cost-efficiency for simple interactions or higher performance for complex customer support. The OpenAI-compatible API makes it easier to integrate these tools into existing workflows without rebuilding your entire infrastructure.

• Lower latency improves the reliability of automated phone systems.
• The v1 Mini model offers a way to scale high-volume interactions while controlling costs.
• Specific model selection allows you to match hardware spend to the complexity of the task.

What to watch

Check your current latency requirements before choosing a model. While v1 Max offers higher accuracy, the v1 Mini may be sufficient for simple tasks like order tracking. You should calculate your expected volume to see if the per-minute pricing aligns with your budget. Test the OpenAI-compatible API to ensure it fits your current technical stack.

Low-latency voice will become a standard utility for internal operations within a year. This shift depends on your ability to map specific tasks to the right model size.