GPT-Live-1 is now in the API
GPT-Live-1 is now available in the API.
Bring ChatGPT’s natural back-and-forth to your app, with voice agents that listen while they speak and work with the models and harness you choose.
GPT-Live-1 is OpenAI’s latest full-duplex voice model designed to bring natural, expressive, and continuous conversation to applications via the API ().
Key features and capabilities include:
- Full-Duplex Communication: The model can listen and speak simultaneously, allowing users to jump in and interrupt smoothly without waiting for a turn to finish (, ).
- Background Noise Handling: It is specifically built to maintain clarity and follow voices even in environments with background noise (, ).
- Backend Delegation: The model acts as a front-end interface, while it pairs with a separate back-end model to handle complex reasoning, tool usage, and task execution (, ).
- Pricing: The service is priced at 5 cents per minute for the front-end voice model, with back-end inference and tools billed separately (, ).
This release enables developers to build sophisticated voice agents that feel more like human-to-human interaction (, ).
- Full-Duplex Communication: Unlike traditional voice bots that require a strict “listen-then-answer” sequence, gpt-live-1 handles interruptions seamlessly and ignores ambient background noise. [3, 4]
- Separation of Layers: The gpt-live-1 model strictly manages the microphone and speaker (the conversation layer). It does not run complex tool calls or reasoning chains itself. [2]
- Backend Delegation: The voice layer automatically hands off complex tasks to a separate backend model (like an OpenAI Responses model or a custom agent) to handle heavy thinking, tool integration, and factual lookups. [2]
- Low Latency: OpenAI benchmarks report a 0.798-second turn-taking latency on its Full Duplex Bench. [3]
- Cost: The front-end voice model is priced at $0.05 (5 cents) per minute ($72 per day for 24/7 runtime). Separate charges apply for any backend reasoning models or tool execution.
- Voices: Shipped with 12 distinct launch voices—Quartz, Ripple, Vesper, Willow, Stone, Gleam, Meridian, Bossa, Tempo, Beacon, Delta, and Cinder.
- Connectivity: Supports integration via WebRTC, WebSocket, and Telephony.
- Safety: Features built-in SynthID watermarking to detect OpenAI provenance signals in generated audio. [2, 3, 4, 5, 6, 7]
- What backend model or agent framework do you plan to use?
- Do you prefer a WebRTC or WebSocket connection setup?
Read more
. Manage external sharing for Gemini Notebook in the Admin console
. Inside the Gemmaverse: Celebrating one billion Gemma downloads
. Personalize the content you see on Search, Discover, and News
. Gemini in Google Sheets is now available on Android devices
. The Gemini desktop app is now available for Windows
. macOS 27 Golden Gate is INCREDIBLE! – First 15 Things
. Take an interactive journey through America’s national parks
. Seamlessly import your team and data from Microsoft to Google Workspace during setup
. Google Workspace Weekly Recap – September 11, 2026
. Set up sharing boundaries for Google Drive with unified data protection rules
for more refer Gemini website click here
for more refer Artificial Intelligence website click here