GPT-Live-1

1 · Family: GPT

OpenAI's conversational voice model released on 8 July 2026, built on a full-duplex architecture that listens and speaks simultaneously. It replaces Advanced Voice Mode in ChatGPT for the paid Go, Plus and Pro tiers, and delegates queries requiring search or reasoning to GPT-5.5 running in the background.

✓ Active✓ Public accessAudioMultimodal📁 GPT

Release date

8 July 2026

🏢OpenAIProducer

Access:HostedDeployment:☁ Cloud

Overview

GPT-Live-1 is OpenAI's new generation of conversational voice models, launched on 8 July 2026 to replace Advanced Voice Mode in ChatGPT. The model introduces a full-duplex architecture: it processes audio input and generates output simultaneously, deciding whether to speak, listen, pause or interrupt many times per second. Users can interject and trail off without the rigid turn-taking rhythm of earlier voice interfaces, and the model signals active listening with brief acknowledgements such as "mhmm" or "got it".

A key architectural element is the separation between the voice model and the reasoning model: when a question requires web search or deeper reasoning, GPT-Live-1 delegates it to GPT-5.5 running in the background while keeping the conversation going. Users can choose an Instant, Medium or High reasoning effort. GPT-Live-1 is the default ChatGPT Voice model for the paid Go, Plus and Pro tiers (GPT-Live-1 mini is the default for the Free tier). ChatGPT ships with nine remastered, predefined voices, and the system includes dedicated voice safeguards described in its System Card.

At launch GPT-Live-1 does not support video or screen sharing and is not yet available through the API — OpenAI has said developer access will follow in the coming months. Some languages (e.g. Hindi in the demo) may sound artificial.

Classification

AudioMultimodal

Family: GPT

Applications

Voice assistant Translation Chatbots

Access & deployment

Hosted

Cloud

Weights: Closed

Key parameters

✓ Tools

📥 Input: text, audio

Technical specification

License

Proprietary

Features:✓ Tool use

Modalities

⬇ Input

textaudio

⬆ Output

textaudio

Capabilities and applications

Native model capabilities

Voice Conversation

Ability to conduct multi-turn real-time voice conversations with context retention and natural speech pacing.

Category: speech

Speech to text

Category: speech

Text to speech

Category: speech

Real-time inference

The model's ability to generate responses with very low latency (>1000 tokens/sec) on specialized inference hardware (e.g. Cerebras WSE), enabling interactive, turn-by-turn collaboration with a human.

Category: coding

Streaming Speech-to-Text

Real-time conversion of speech to text with immediate output as the speaker is talking.

Category: speech

Live Translation

Real-time speech translation between multiple languages without interrupting the audio stream.

Category: speech

Multilingual

Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.

Category: language

Application domains

Voice assistant Translation Chatbots