Breaking

OpenAI Releases Full-Duplex Voice Model GPT-Live-1

July 8, 2026 at 19:23 EDT

Foundation Models
Media Generation

On July 8, 2026, OpenAI began rolling out new models that overhaul ChatGPT's voice experience: "GPT-Live-1" (for paid plans) and "GPT-Live-1 mini" (for free plans). Both models adopt a full-duplex architecture in which the AI listens simultaneously while the user is speaking, allowing it to respond with backchanneling and interruptions.

July 8, 2026 · OpenAI

ChatGPT Voice Learns to Listen and Talk at the Same Time

New full-duplex voice models — GPT-Live-1 for paid plans and GPT-Live-1 mini for free — let the AI hear you while it speaks, handling interruptions and back-channel cues so conversations feel closer to human.

new voice models: GPT-Live-1 (paid) + mini (free)

consumer plans covered: Free · Go · Plus · Pro

100%

full-duplex — listens & speaks simultaneously

Turn-based vs Full-duplex

Old mode waited for you to finish. The new model overlaps listening and speaking.

LISTEN

↓

then SPEAK

Advanced Voice Mode

sequential turns

LISTEN + SPEAK

at the same time

handles interruptions

GPT-Live-1

overlapping / real-time

What it can do

Streaming voice + text side by side
Web search & memory
Visual display via widgets
Mix voice, text & images in one chat

Not yet

Video & screen share (still via old mode)
Business / Enterprise / Edu workspaces
Confirmed pricing & benchmarks

Praised

"Like talking to a human." Smooth interruptions, real-time translation and a clean flowing UI — good for language learning, discussing news and brainstorming.

Criticized

Emotional expression and accents still weak. No enterprise or video support. As conversations grow more human-like, safety and regulation concerns rise.

OpenAI is positioning voice as a core ChatGPT experience — betting on naturalness and low latency as its edge over rival voice features like Gemini and Claude.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.