Decagon Alto uses a separate model to decide when voice agents should speak
Decagon's Alto model manages turn-taking and response timing in a duplex voice-agent architecture; the company reports average latency of about 700 ms.
AT A GLANCE
Key points
- Separate model for turn-taking and response timing
- Provider reports about 700 ms average response
- Tested on more than 90 million simulated calls
What is changing?
Decagon introduced Alto, a model for voice agents, on November 11. In the company's duplex architecture, Alto manages the conversational layer: it decides when to listen, when to speak, and whether to answer or wait for confirmation. Decagon says Alto responds in about 700 milliseconds on average without relying on filler words.
The company says it started from a small open-weight model, fine-tuned it with simulated calls and reinforcement learning, tested it on more than 90 million simulated calls, and then ran it in shadow mode on live traffic. According to Decagon, Alto increased live-call resolution by five percentage points without significant changes to the logic layer. This provider-reported result still needs validation across different industries and call types.
Explore the original sources
Source date: 11 November 2026.
Published by AI Radar: 11 Nov 2026, 21:12 (Istanbul, UTC+3).
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…