Audio Input
AI Models & Assistants / Knowledge library
GPT-Live 1
Our premier model for natural, expressive voice conversations with smooth interruption handling.
Why learn GPT-Live 1?
Offers the documented OpenAI model capabilities through hosted.
The problem it solves
Provides model inference for supported audio / text inputs.
Where it came from
GPT-Live 1: Our premier model for natural, expressive voice conversations with smooth interruption handling. Published aliases: gpt-live-1. Input: audio, text. Output: audio, text. Checked 2026-10-13. Model capabilities are distinct from tools available in a particular agent.
Build your understanding
Work through these exercises in order. Try the task before opening a reference.
- Explain what GPT-Live 1 does.
- Identify a use case for GPT-Live 1.
- Describe a limitation of GPT-Live 1.
- Name a compatible tool for GPT-Live 1.
- Compare GPT-Live 1 against an alternative.
Documentation and sources
- Primary source / model card (developers.openai.com)
Catalog research notes
- See the AI Workbench for source dates, pricing evidence and configuration details.
Connect the ideas
Built on
Real World Example
Dynamic Concierge for High-End Hospitality
Luxury hotel chains managing high-volume guest inquiries via voice channels.
The business problem
Traditional IVR systems and standard text-based bots fail to handle natural interruptions, leading to guest frustration when they attempt to clarify requests mid-sentence or correct information during a booking process.
How this topic solves it
GPT-Live 1 provides a full-duplex voice interface that allows guests to speak naturally. Because the model supports smooth interruption handling, it can process a guest's correction instantly without waiting for a pre-recorded prompt to finish, maintaining a fluid, human-like interaction.
- Initialize a session using the v1/live/sessions endpoint to establish the full-duplex connection.
- Configure the backend agent to handle specific hotel operations like room service or concierge requests.
- Implement function calling to allow the model to query internal reservation databases during the live audio stream.
- Monitor concurrent session limits to ensure the infrastructure scales with peak check-in hours.
The flow shows how the guest interacts with the model and how the model delegates tasks.
Full-duplex processing
Logic and Tools
Data storage
- Guest → interrupts/speaks → GPT-Live 1
- GPT-Live 1 → delegates task → Backend Agent
- Backend Agent → queries/updates → Reservation DB
- Backend Agent → returns result → GPT-Live 1
Expected business outcome
Guests experience a more responsive and empathetic service interaction, reducing the perceived friction of automated booking systems.
Limitations and checks
The primary trade-off is the reliance on real-time full-duplex processing, which requires stable network conditions; developers should check for latency spikes in the audio stream to determine if the connection quality is degrading.