OpenAI launched GPT‑Live‑1 in the API on September 10, 2026, giving developers a full-duplex model for voice-enabled applications and business workflows. The front-end voice layer is available at $0.05 per minute and can be paired with a developer-selected backend model and agent harness.
GPT‑Live‑1 listens and speaks through a single model, allowing it to respond to interruptions and acknowledgements while deeper reasoning proceeds in the background. OpenAI positions this design as a simpler alternative to chained speech-to-text, reasoning and text-to-speech systems, whose handoffs can add latency or lose conversational context. The model can delegate reasoning and tool calls to a backend text model such as GPT‑6 Astra or a third-party model.
Developers can use system prompts to shape an agent’s tone, pace and conversational style. OpenAI also says the model handles silence and background noise without unnecessarily interrupting or narrating each step, retains context across extended interactions, and supports full-duplex telephone agents for uses including reservations and customer support. It provides ASR transcripts and response text, supports keyword biasing and alphanumeric understanding, and offers native turn detection despite not being a turn-based model.
In OpenAI’s 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. OpenAI also reported that GPT‑Live‑1 paired with GPT‑6 Astra at medium reasoning effort ranked first on Tau3, an evaluation of end-to-end voice-agent tasks. In early evaluations by language-learning company Speak, the model cut interruptions during learners’ thinking pauses by almost 80% compared with previous turn-based systems.
OpenAI cited early deployments and tests across Yelp, Speak, Fin and Cognition. Tony Stoyanov, identified as a co-founder and chief technology officer, said replacing a cascaded implementation with GPT‑Live‑1 simplified the code base by 80% and removed 23,000 lines of code. Yelp Chief Technology Officer Alex Levy said integrations in Yelp Host and Hatch improved turn-taking and accuracy over the company’s traditional voice architecture, alongside improvements in call-handling rates.
The release expands OpenAI’s selection of real-time voices across accents, dialects and languages, with further voice and language availability planned over the coming months. Custom voice access requires contacting sales about eligibility and the request process. OpenAI also offers Presence, which uses GPT‑Live‑1 for enterprise voice agents that can use company systems, take approved actions and escalate interactions to people.