GPT-Live 1 Launch: A Full-Duplex Conversational Voice Model in the API

This session introduces GPT-Live 1, a full-duplex voice model now available in the API, demonstrating how it supports natural, expressive conversation while listening and speaking at the same time. It also outlines how the model pairs with a back-end model for reasoning and tool use, along with its pricing for production deployment.

■ Model Capabilities
- Full-duplex operation, listening and speaking simultaneously
- Handles background noise and interruptions smoothly

■ Architecture
- Pairs with a back-end model for reasoning and tools
- Used to delegate actions to a robot and display while conversation continues

■ Availability and Pricing
- Launched in the API for production conversational voice
- Front-end model priced at 5 cents per minute
- Back-end inference and tool services billed separately

This overview suits developers and builders evaluating conversational voice for production, offering a concise look at the model's interaction capabilities, its pairing with a back-end model, and its pricing structure.