GPT-Live is OpenAI's new generation of voice models, announced 8 July 2026. It uses a full-duplex architecture — listening and speaking at the same time — so conversations flow like a phone call rather than walkie-talkie turns. It now powers ChatGPT Voice.

How much does GPT-Live cost?

There is no separate charge in ChatGPT — GPT-Live-1 is the default voice model for Go, Plus and Pro subscribers, and free accounts get GPT-Live-1 mini. API pricing is TBD; OpenAI has announced API availability as coming soon without publishing rates.

How smart is GPT-Live?

OpenAI calls it its smartest voice model yet. For questions needing web search, deeper reasoning or complex work, GPT-Live delegates to OpenAI's latest frontier model behind the scenes — GPT-5.5 at launch — and folds the result back into the conversation while keeping the dialogue flowing.

When can developers use GPT-Live?

No date yet. OpenAI says GPT-Live models are coming to the API 'soon' for developers and enterprises. Until model IDs and a rate card are published, treat all API details as TBD.

GPT-Live (GPT-Live-1 & GPT-Live-1 mini)

Q: Is GPT-Live a model or a ChatGPT feature?

It is a named model family — GPT-Live-1 and GPT-Live-1 mini — that ships inside ChatGPT Voice. OpenAI says the models are coming to the API 'soon' for developers and enterprises, but no API model IDs or per-token pricing have been published yet.

Q: What does full-duplex mean?

The model continuously processes your incoming audio even while generating its own speech — like both parties on a phone call talking and listening simultaneously. GPT-Live can back-channel ('mhmm', 'yeah'), handle quick interruptions, or stay quiet while you think, instead of waiting for a silence gap.

OpenAI's new voice model family, announced 8 July 2026 — full-duplex models that listen while they speak, now rolling out as the engine behind ChatGPT Voice. API and pricing: not yet published.

Compiled by AI Model Hub · Last updated 9 July 2026 · Sourced from OpenAI's announcement and press coverage · API-stage specs still TBD

GPT-Live is OpenAI's new generation of voice models, announced on 8 July 2026. Two models — GPT-Live-1 (default for Go, Plus and Pro subscribers) and GPT-Live-1 mini (free tier) — use a full-duplex architecture that listens and speaks simultaneously, and delegate harder questions to GPT-5.5 behind the scenes. Rolling out globally in ChatGPT over days; API availability announced as "soon", pricing TBD.

Status: live in ChatGPT · API TBD

GPT-Live is a genuine model family, not just a feature toggle — but OpenAI has published no API model IDs, rate card or context specs. This page tracks what's confirmed; API fields stay TBD until OpenAI publishes them.

On this page

What's confirmed
Full-duplex, explained
Delegation to GPT-5.5
Rollout & tiers
API status
FAQ

GPT-Live: what's confirmed so far

Maker	OpenAI
Announced	8 July 2026
Type	Voice (speech-to-speech) model family — full-duplex
Variants	GPT-Live-1 (Go, Plus, Pro default) · GPT-Live-1 mini (free tier)
Status	Rolling out globally in ChatGPT — iOS, Android and web, over days from launch
Backend delegation	Hands off hard questions to OpenAI's latest frontier model — GPT-5.5 at launch
API availability	Announced as coming "soon" — no date, no model IDs
API pricing	TBD — not published
Context window	TBD — not published
Benchmarks	TBD — no public model card with figures yet

Calculate LLM API costs → This model has no published USD per-token rate card, so it is not in the calculator — compare 16 published rate cards instead.

Full-duplex, explained

Every previous ChatGPT voice mode worked like a walkie-talkie: it waited for a silence gap, decided you were done, then answered. GPT-Live's full-duplex architecture processes your incoming audio continuously while it is generating its own speech — the way both parties on a phone call talk and listen at once.

In practice that means GPT-Live can back-channel ("mhmm", "yeah") to show it's following, handle rapid back-and-forth without resetting, absorb interruptions mid-sentence, or simply stay quiet when you pause to think instead of jumping in. OpenAI calls it its smartest voice model yet, and it replaces the previous voice pipeline as the engine of ChatGPT Voice.

Delegation to GPT-5.5

GPT-Live is not trying to be the smartest model in the building — it's the conversational front-end. For questions that need web search, deeper reasoning or longer work, it delegates to OpenAI's latest frontier model behind the scenes and folds the result back into the conversation when ready, keeping the dialogue flowing while it waits. At launch the background model is GPT-5.5.

Architecturally this is the same pattern as fast/deep model pairs elsewhere: a low-latency specialist handles the interaction loop, a heavyweight handles the hard steps.

Rollout & tiers

GPT-Live-1 — becomes the default voice model for ChatGPT Go, Plus and Pro subscribers.
GPT-Live-1 mini — the version for free accounts.
Where — iOS, Android and ChatGPT.com, rolling out globally "over the next few days" from 8 July 2026.

API status: announced, not shipped

OpenAI says it plans to bring the GPT-Live models to the API "soon" for developers and enterprises — but as of 9 July 2026 there are no API model IDs, no per-token or per-minute rates, and no context specs in OpenAI's pricing or model docs. Until that changes, GPT-Live is a ChatGPT-product model family with a pending API, and every API field on this page stays TBD. The closest shipping developer option today remains OpenAI's gpt-realtime family.

Read OpenAI's announcement

Frequently asked questions

What is GPT-Live?

OpenAI's new generation of voice models, announced 8 July 2026 — full-duplex models that listen and speak at the same time, now powering ChatGPT Voice.

Is GPT-Live a model or a ChatGPT feature?

A named model family — GPT-Live-1 and GPT-Live-1 mini — shipping inside ChatGPT Voice. API access is announced as coming "soon" but has no published IDs or pricing yet.

How much does it cost?

Nothing extra inside ChatGPT — GPT-Live-1 is the default for Go/Plus/Pro, free accounts get GPT-Live-1 mini. API pricing is TBD.

What does full-duplex mean?

The model keeps processing your audio while it speaks — enabling back-channels, natural interruptions and phone-call-like flow instead of walkie-talkie turns.

How smart is it?

OpenAI calls it its smartest voice model yet; for hard questions it delegates to GPT-5.5 behind the scenes and brings the answer back into the conversation.

When can developers use it?

No date yet — OpenAI says the models are coming to the API "soon" for developers and enterprises.

Sources & further reading: