OpenAI GPT-Live-1.1: New Voice Model Improves Transcription and Accents, Adds Cheaper Standby

OpenAI released GPT-Live-1.1 in its Live API on 19 November 2026. The full-duplex voice model transcribes speech more accurately and keeps a more consistent accent across languages, and WebRTC sessions can now pause in standby at $0.02 per minute.

Christopher Ort
By Christopher Ort · 3 min read
Share
A headset with a microphone on a desk next to a laptop in a quiet office.
Image: AI-generated, i10X
On this page
  1. Quick Take
  2. In plain English
  3. What happened
  4. Background
  5. How it works
  6. Why it matters
  7. What's still open
  8. Sources
  9. i10x Perspective

OpenAI has released GPT-Live-1.1, an updated version of its full-duplex voice model for developers. According to the OpenAI API changelog, the new model improves pronunciation, accent consistency and transcription in multilingual conversations, and both GPT-Live versions now get a cheaper standby mode for idle WebRTC sessions.

i10X Quick Take

Quick Take

  1. What’s new: GPT-Live-1.1 is available in the Live API under the model ID gpt-live-1.1, listed in the changelog on 19 November 2026.
  2. Better speech: OpenAI says it “transcribes speech more accurately and maintains a more consistent accent and voice across languages.”
  3. Full duplex: the model “can listen while it speaks” and hands reasoning or tool work to a backend model.
  4. Pricing: $0.05 per minute of active session and $0.02 per minute in standby; backend model and tool usage are billed separately.
  5. Standby: WebRTC sessions on GPT-Live-1 and 1.1 can pause inference during silence and resume when someone speaks.

In plain English

GPT-Live is the engine developers use to build apps you can talk to, like a phone support line or a voice assistant inside an app. It can listen and talk at the same time, so you can interrupt it like a person. Version 1.1 understands words more reliably and keeps the same accent when it switches languages. A new standby mode means apps pay less while nobody is speaking.

What happened

OpenAI added GPT-Live-1.1 to its API changelog on 19 November 2026: “GPT-Live-1.1 improves pronunciation, accent consistency, and transcription in multilingual voice conversations.” Developers switch by setting gpt-live-1.1 in an existing Live session configuration.

The same entry adds standby: developers can “configure standby for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions to pause inference during inactivity and resume when someone speaks.” The lower $0.02 per minute rate applies while inference is paused.

Background

GPT-Live is OpenAI’s voice model family for the Live API, which we covered when OpenAI introduced GPT-Live. Since mid-October OpenAI has added outbound phone calls over SIP for all organizations, custom voices created from text prompts and nine new voices in languages including German, French, Hindi, Japanese and Korean. Rivals are moving fast too, with Grok Voice 3.0 competing on latency and flat per-minute pricing.

How it works

  • Two layers: GPT-Live handles the spoken conversation, while a backend does the thinking. With Responses delegation, GPT-Live calls a Responses model the developer picks; with client delegation, the app runs its own agent or service and returns the result.
  • Plain-language rules: developers “describe in plain language when the model should hand a task to your backend for reasoning or tool use.”
  • Connection: sessions start through POST /v1/live/sessions, with WebRTC for browser and mobile apps or SIP for phone calls. Connected calls are limited to two hours.
  • Inputs and outputs: audio and text in, audio and text out, with streaming and function calling. The model is not available through Chat Completions, Responses or the older Realtime endpoint.
  • Limits: GPT-Live-1.1 and GPT-Live-1 have separate concurrent-session limits.

Why it matters

For developers building voice agents, transcription errors on names, numbers and mixed languages are a main reason calls fail. A model that hears these more accurately reduces the need for workarounds such as asking callers to spell everything.

For companies running call centers or in-app assistants, standby billing at $0.02 per minute lowers the cost of sessions that stay open while users read, think or wait on hold.

For multilingual markets, a consistent accent and voice across languages matters for brands that want one recognizable assistant in many countries.

What’s still open

  • OpenAI has not published benchmark numbers comparing GPT-Live-1.1 with GPT-Live-1.
  • Whether ChatGPT’s own voice mode uses the same update was not stated in the developer documentation.
  • How long GPT-Live-1 remains available alongside the new version.

Sources

  • OpenAI API changelog, entry for 19 November 2026
  • OpenAI API documentation, “GPT-Live-1.1” model page and pricing
  • OpenAI API documentation, “Getting started with GPT-Live” and “Delegation and tools in GPT-Live”

i10x Perspective

Voice is turning into the front door for AI agents, and the small details decide whether people trust it: a misheard digit, an accent that drifts mid-sentence. GPT-Live-1.1 is an incremental release, but standby pricing shows OpenAI is optimizing for always-on voice apps. More model news in Models.

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Timeline

What happened before

  1. Release

    GPT-6.1 Terra

    1.05M context · $1 / $5 per 1M tokens

    openai.com (source, opens in a new tab)
  2. Acquisition

    OpenAI acquires Orchestra Bio

    Acquires Orchestra Bio

    openai.com (source, opens in a new tab)
  3. Release

    GPT-6.1 Sol

    1.05M context · $2 / $10 per 1M tokens

    developers.openai.com (source, opens in a new tab)
  4. Release

    GPT-6 Sol & Luna

    1.05M context · $2 / $10 per 1M tokens

    Two GPT-6 reasoning models. Price shown is GPT-6 Sol (≤272K prompt); GPT-6 Luna costs $0.10 / $0.50.

    developers.openai.com (source, opens in a new tab)
  5. Release

    GPT-6 Astra

    1.05M context · $10 / $50 per 1M tokens

    OpenAI's most capable model for reasoning, coding, computer use and research.

    openai.com (source, opens in a new tab)
Full OpenAI timeline ›

Editor's picks

  1. What Comes After AGI? Google DeepMind Maps Four Paths to Superintelligence
  2. AMD and Unsloth Bring NPU Support to Unsloth Desktop, Letting 100+ AI Agent Apps Run Locally on Ryzen AI
  3. Perplexity Brain Now Writes Its Own Agent Skills, With 80,000 Created for 20,000 Users
  4. OpenAI Performance Factory: Codex Agents Ran 1,700 Investigations to Keep ChatGPT Fast
  5. OpenAI GPT-Live-1.1: New Voice Model Improves Transcription and Accents, Adds Cheaper Standby