What is OpenAI GPT-Live-1?

GPT-Live-1 is OpenAI’s dedicated real-time speech-to-speech foundation model available through the OpenAI Realtime API. It provides bidirectional, full-duplex audio streaming with simultaneous listening and speaking capabilities.

How does GPT-Live-1 handle human conversational interruptions?

Because GPT-Live-1 ingests audio continuously while generating output packets, its acoustic decoder detects user vocal energy in real time. It halts playback within 80ms of genuine user speech, clearing client buffers without audio clipping.

What is the cost of running GPT-Live-1 in production?

OpenAI prices the voice layer at $0.05 per active audio minute, equivalent to $3.00 per hour of continuous bidirectional streaming. This replaces separate billing for ASR, text LLM inference, and external TTS synthesis.

How does GPT-Live-1 differ from the previous gpt-4o-realtime-preview?

GPT-Live-1 is optimized specifically for conversational flow and speech fidelity. It provides tighter background noise filtration, lower turn-taking latency (down from 450ms to 290ms), lower per-minute compute costs, and native support for asynchronous model handoffs.

Can GPT-Live-1 run over standard telephony and WebRTC?

Yes. Developers can connect to GPT-Live-1 directly via WebRTC data and audio channels using ephemeral session tokens or through integration frameworks like LiveKit Agents, which bridge directly to SIP trunks (Twilio, Telnyx) for phone calls.

OpenAI GPT-Live-1 Deep Report: Realtime Voice API, Benchmarks & $0.05/Min Pricing

A technical analysis of OpenAI’s GPT-Live-1 API release. Covers native speech-to-speech neural streaming, full-duplex conversational turn-taking, 80.1% interactivity benchmarks, $0.05-per-minute audio pricing, LiveKit WebRTC integration, and enterprise deployments across Yelp and Hatch.

# OpenAI Releases GPT-Live-1: Realtime Full-Duplex Voice API, $0.05/Min Economics, and Sub-300ms Turn-Taking OpenAI launched **GPT-Live-1** in its developer API, opening the full-duplex conversational voice system previously restricted to ChatGPT Advanced Voice Mode. The release provides developers with a dedicated, low-latency foundation model designed for bidirectional audio streaming. GPT-Live-1 processes acoustic tokens natively. By eliminating the traditional three-hop pipeline of speech recognition, text inference, and speech synthesis, the model achieves median end-to-end audio latencies of **280ms to 320ms**. It handles conversational interruptions, filters background kitchen and street noise, and executes dynamic handoffs to reasoning models for heavy compute tasks. OpenAI priced the voice layer at **$0.05 per minute** ($3.00 per active hour). Enterprise adopters, including Yelp and Hatch, deployed the model across customer-facing phone networks and consumer hardware follo...