OpenAI GPT-Live Replaces Voice Pipeline With Full-Duplex Model, API Pricing Unavailable
OpenAI's July 8, 2026 GPT-Live release replaces the prior three-stage speech-to-text, language model, text-to-speech chain with a single neural model that listens and speaks simultaneously, delegating complex queries to GPT-5.5 in the background without pausing the conversation. The product is live for ChatGPT subscribers across Free through Pro tiers but has no API release date and no published per-token pricing; the closest available proxy is the existing Realtime API at $32 per million audio input tokens and $64 per million audio output tokens. No head-to-head benchmark of GPT-Live against Google Gemini Live exists yet, and OpenAI has not published latency figures in milliseconds for the new model.