OpenAI announced GPT‑Live‑1 in the API on September 10, 2026, a model designed to listen and speak simultaneously to support more natural, real‑time voice interactions. The release emphasizes reduced latency and a simpler voice layer by combining listening and speaking in a single model rather than chaining separate STT–LLM–TTS components.
Developer features and integration
GPT‑Live‑1 handles interruptions and background noise while allowing developers to steer tone, pace, and conversational style through system prompts. It can delegate deeper reasoning and tool calls to backend text models such as GPT‑6 Astra, Luna, Terra or third‑party services, and natively provides ASR transcripts, response text, alphanumeric understanding and keyword biasing. Although not turn‑based, the model supports explicit turn detection for workflows that need discrete turns.
The model is positioned for telephony and other real‑time applications, with examples ranging from restaurant reservations to customer support. A code example in the announcement shows how an application can pass conversation context to a Codex thread and return answers to GPT‑Live‑1; connection setup and delegation handling were not included in that excerpt.
Evaluations, customer reports and availability
In evaluations, GPT‑Live‑1 improved Full Duplex Bench performance by 30 percentage points over GPT‑Realtime‑2.1, with gains in turn‑taking latency and interactive behavior. Paired with GPT‑6 Astra at medium reasoning effort, it ranked #1 on Tau3 for end‑to‑end voice‑agent tasks. Early user testing by Speak reported the model cut interruptions by almost 80% compared with prior turn‑based systems.
Customers quoted in the announcement reported operational impacts: Tony Stoyanov, Co‑Founder & CTO, said, “Compared to our cascaded build, GPT‑Live‑1 simplified our code base by 80% and removed 23K lines of code.” Alex Levy, Chief Technology Officer at Yelp, said adding GPT‑Live‑1 improved turn‑taking and accuracy and produced “meaningful improvements in call handling rates.”
GPT‑Live‑1 expands available real‑time voice options (Quartz, Ripple, Vesper, Willow, Stone, Gleam, Meridian, Bossa, Tempo, Beacon, Delta, Cinder) and is available in the API today at $0.05 per minute for the front‑end voice layer. For custom voice access or enterprise deployment options such as OpenAI Presence, contact sales or an account director for eligibility and requests.
Original source: OpenAI News