OpenAI split a voice model’s brain. Then one team deleted 23,000 lines of code.
Building an AI voice agent has always been clunkier than it seems. Most voice agents are really a chain of The post OpenAI split a voice model’s brain. Then one team deleted 23,000 lines of code. appeared first on The New Stack .
OpenAI unveiled GPT-Live-1, a new voice model in its API, allowing developers to utilize full-duplex voice architecture for AI conversations. This eliminates the need for developers to manage a chain of systems, as GPT-Live-1 can handle the conversation and offload heavier processing to other models in the background. This feature enables continuous dialogue without awkward pauses, even when a more complex model like GPT-6 Astra is required.
The company reports a 30% performance improvement over GPT-Realtime-2.1 on Full Duplex Bench and outperforms on the τ⁽³⁾-benchmark when paired with GPT-6 Astra at medium reasoning. Early adopters, such as EliseAI and Speak, have seen significant code reduction and improved conversation flow. GPT-Live-1 pricing is $0.05 per minute, with additional costs for any backend models used.
The shift to this model allows developers to focus on user experience rather than building separate conversation systems.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
