OpenAI's GPT-Live-1.1 voice model reaches developers, and charges less while the caller hunts for their booking number
By Brett Gavaghan, Founder of Adaptd ·
OpenAI released GPT-Live-1.1 in its API on 20 November, the voice model behind ChatGPT's voice mode. A new Standby setting bills quiet stretches of a call at US$0.02 a minute, down from US$0.05.
The short version
- GPT-Live-1.1 is OpenAI's voice model for building AI phone and voice agents, now available in its API.
- It reached developers on 20 November 2026, with a Standby mode billed at US$0.02 a minute against US$0.05 for active talk.
- If you run or are pricing a voice agent, ask your provider whether it uses GPT-Live and whether Standby will cut your costs.
Every phone call has a stretch where nothing happens while someone looks for a reference number. Until now, a voice agent billed you full price for that silence.
What is GPT-Live-1.1?
GPT-Live-1.1 is the OpenAI voice model that has run voice conversations in ChatGPT Plus and Pro since 15 October 2026, and developers can now build it into their own apps. OpenAI says it keeps accents more natural and consistent across languages, catches details like letters and numbers more reliably and follows instructions better.
That last part matters more than it sounds. An agent that mishears one letter of an email address books the appointment for someone who doesn't exist.
What does it cost?
GPT-Live charges US$0.05 a minute while the model is active, and the new Standby setting drops that to US$0.02 a minute during silence. Developers set an idle timeout of 5 to 60 seconds, after which the model pauses but the call stays open, and it picks up again when someone speaks.
Standby works with GPT-Live-1 and GPT-Live-1.1. Any backend model and tool use in the call is billed separately, so the per-minute rate isn't the whole bill.
Should your business build a voice agent with it?
Your business doesn't need to build anything to benefit, because most small businesses buy a voice agent from a provider instead of coding one.
My take: if you already pay for an AI receptionist or are getting quotes, ask two questions. Which voice model does it run on, and does the price drop when the line goes quiet? Then ring it yourself and spell out a tricky surname. If it gets that right, it's earning its keep.
You wouldn't pay a receptionist double for the minutes they spend on hold. Your phone bot shouldn't charge you full rate either.
Wondering what this means for your business?
That's exactly what the discovery call is for.