GPT-Live-1

NewOpenAIvoicespeech-to-speech

OpenAI's full-duplex speech-to-speech voice model, released July 8, 2026 alongside a lighter GPT-Live-1 mini. It has replaced Advanced Voice Mode as ChatGPT's default voice experience.

GPT-Live-1 strengths

  • Full-duplex: listens and speaks simultaneously
  • Natural interruptions and backchanneling
  • Live translation
  • Delegates web search and reasoning to GPT-5.5 for harder tasks

Pricing & context

Context windowNot yet published
Input price /1MAPI coming soon
Output price /1MAPI coming soon
Modalitiesspeech-to-speech (voice)

There is no API pricing to model yet: OpenAI has announced API access as "coming soon" but has not published rates. Today, GPT-Live-1 is used through ChatGPT — the full model on Go ($8/mo), Plus ($20/mo) and Pro plans, with GPT-Live-1 mini rolling out to Free users.

A voice model, not a text LLM

GPT-Live-1 is a speech-to-speech voice model, not a text LLM, so it is not directly comparable to models like GPT-5.5 on text benchmarks. Instead of competing on reasoning scores, it delegates: for harder tasks it hands web search and reasoning to GPT-5.5 behind the scenes while it handles the live conversation.

When to choose GPT-Live-1

GPT-Live-1 replaced Advanced Voice Mode as the default voice experience in ChatGPT, so most users will meet it there rather than choose it. It is built for natural spoken conversation — interrupting mid-sentence, backchanneling, and live translation — rather than text work. Developers wanting to build on it will need to wait for the API, which OpenAI has announced but not yet priced.

GPT-Live-1 FAQ

What is GPT-Live-1?

GPT-Live-1 is a full-duplex speech-to-speech voice model released by OpenAI on July 8, 2026. It listens and speaks simultaneously, supporting natural interruptions, backchanneling, and live translation, and it has replaced Advanced Voice Mode as ChatGPT's default voice experience.

How much does GPT-Live-1 cost?

GPT-Live-1 is included in ChatGPT plans: the full model on Go ($8/mo), Plus ($20/mo), and Pro, while GPT-Live-1 mini is rolling out to Free users. API access is announced as coming soon, and no API pricing has been published yet.

Is GPT-Live-1 comparable to text models like GPT-5.5?

No. GPT-Live-1 is a voice model, not a text LLM, so it is not directly comparable on text benchmarks. For harder tasks it delegates web search and reasoning to GPT-5.5 behind the scenes.

What is GPT-Live-1 mini?

GPT-Live-1 mini is the lighter version of GPT-Live-1, also released July 8, 2026. It is becoming the default ChatGPT voice experience for Free users, with the rollout still in progress.

Tools that use GPT-Live-1

Other models

All models →
01Claude Fable 5Anthropic$10.00
02GPT-5.5OpenAI$5.00
03Claude Opus 4.8Anthropic$5.00
04Gemini 3.1 ProGoogle$2.00 (under 200K; $4.00 above)