All vikis
007 First Light
AI
Aion 2
Ananta
ARC Raiders
ArcheAge Chronicles
Call of Duty: Modern Warfare 4
Chrono Odyssey
Civilization 3
Civilization 4
Claude
Crimson Desert
Deadlock
Diablo 4
Diablo 5
Enshrouded
Fable
Final Fantasy Resonance
Football Manager 27
Forza Horizon 6
God of War
Grand Theft Auto 6
Heroes of Might and Magic 3Realtime API
The OpenAI Realtime API streams speech in and speech out with low latency over WebRTC or WebSocket, using models such as gpt-realtime-2.1 and the full-duplex GPT-Live 1 ($0.05 a minute), as of Nov 6, 2026. It also handles live transcription and speech-to-speech translation.
The OpenAI Realtime API lets an app hold a spoken conversation with a model: audio streams in and the model answers in audio, with tool calls along the way, over WebRTC in the browser or WebSocket on a server. As of Nov 6, 2026 the main models are gpt-realtime-2.1, a reasoning speech-to-speech model priced at $32 per 1M audio input tokens and $64 per 1M audio output tokens, its cheaper mini sibling, and GPT-Live 1, a full-duplex voice model that listens and speaks at once for $0.05 a minute. The Realtime API has been generally available since Aug 28, 2025. It is the developer side of what ChatGPT users know as Voice Mode; for file transcription and text to speech, see Whisper and Audio Models.
Quick facts
| Generally available | Aug 28, 2025; the Beta was removed May 12, 2026 |
| Connections | WebRTC (browser) and WebSocket (server) |
| Main model | gpt-realtime-2.1, released Jul 6, 2026 |
| Full-duplex voice | gpt-live-1, generally available Sep 10, 2026 |
| GPT-Live price | $0.05 per minute, billed per second |
| gpt-realtime-2.1 audio | $32 in / $0.40 cached / $64 out per 1M tokens, as of Nov 6, 2026 |
| gpt-realtime-2.1-mini audio | $10 in / $0.30 cached / $20 out per 1M tokens |
| Live translation | gpt-realtime-translate, about $0.034 a minute |
Which realtime model should you use?
OpenAI's models page lists six realtime models as of Nov 6, 2026:
| Model | What it is | Released |
|---|---|---|
gpt-live-1 |
Full-duplex voice; reasoning and tools handed to a backend agent | GA Sep 10, 2026 |
gpt-realtime-2.1 |
Reasoning speech-to-speech model with tool use | Jul 6, 2026 |
gpt-realtime-2.1-mini |
Smaller, cheaper 2.1 | Jul 6, 2026 |
gpt-realtime-2 |
The previous reasoning model, same price as 2.1 | May 7, 2026 |
gpt-realtime-translate |
Streaming speech-to-speech translation | May 7, 2026 |
gpt-realtime-1.5 |
"The best voice model for audio in, audio out" | Feb 23, 2026 |
For most voice agents, gpt-realtime-2.1 is the default OpenAI's own Realtime guide uses in its code (Realtime guide). Choose gpt-realtime-2.1-mini when cost matters more than reasoning, and GPT-Live 1 when the conversation must feel natural, with interruptions and both sides talking at once.
gpt-realtime-2.1
OpenAI says gpt-realtime-2.1 is better than GPT-Realtime-2 at letters and numbers, silence and background noise, and interruptions (model page). Its specifications:
- Context window 128,000 tokens, maximum output 32,000 tokens.
- Knowledge cutoff Sep 30, 2024, much older than the GPT-6 text models, so pair it with tools or a backend for current facts.
- Input text, audio and images; output text and audio.
- Endpoint
v1/realtimeonly. - Rate limits: Build 400 requests and 200,000 tokens a minute; Launch 10,000 and 4,000,000; Grow 20,000 and 15,000,000.
Text and image tokens cost less than audio: text $4.00 in, $0.40 cached, $24.00 out per 1M tokens, and image input $5.00.
GPT-Live 1
GPT-Live 1 is a full-duplex model: it listens and speaks at the same time and hands reasoning and tool use to a separate backend agent (model page). OpenAI describes it as its "premier model for natural, expressive voice conversations with smooth interruption handling" (models).
- Endpoint:
v1/live/sessionsonly (model page). - Price: $0.05 a minute, billed per second and not rounded up; the backend model and tools are billed separately.
- Concurrent sessions: 50 on Build, 300 on Launch, 500 on Grow; the Free tier is not supported.
- Knowledge cutoff: Jul 31, 2025.
- Voices: nine new voices and prompt-to-voice (describing a voice in words) were added on Oct 16, 2026.
ChatGPT's own Live voice runs on GPT-Live-1.1 and GPT-Live-1.1 Quip (Help Center); OpenAI published the GPT-Live-1.1 system card on Oct 15, 2026. The API model page read for this article lists gpt-live-1.
How to connect
- Pick the transport. Browsers and mobile apps use WebRTC; servers use WebSocket. WebRTC sessions are set up through
/v1/realtime/calls(Realtime guide). - Never ship your API key to the client. Your server calls
POST /v1/realtime/client_secretsto create an ephemeral credential for the browser or mobile app. - Open the session from the front end with that secret, choosing the model (for example
gpt-realtime-2.1) and instructions. - Stream audio both ways, and handle tool calls as they arrive.
Phone calls over SIP were part of the brief for this page, but the Realtime guide read on Nov 6, 2026 does not mention SIP, so this page makes no claim about it.
Transcription and translation in real time
- Live captions:
gpt-live-transcribeandgpt-realtime-whisper, each about $0.017 a minute. The guide sends live captions to "Live transcription" and recorded audio to "File transcription" (Realtime guide). - Translation:
gpt-realtime-translatetranslates speech to speech as it streams, about $0.034 a minute. The old/v1/audio/translationsendpoint shuts down Feb 26, 2027, and translation moves tov1/realtime/translations(deprecations).
Older realtime and audio models
gpt-realtime-1.5 and the original gpt-realtime cost $32 audio in and $64 out, with text at $4 and $16. The chat-style audio models gpt-audio-1.5 and gpt-audio cost the same for audio; gpt-audio-mini costs $10 and $20. The text to speech models tts-1, tts-1-hd and gpt-4o-mini-tts shut down on Jan 6, 2027, and OpenAI names gpt-realtime-2.1-mini as their replacement.
History
- Aug 28, 2025: Realtime API generally available (changelog).
- Oct 6, 2025 (DevDay): GPT-Realtime mini and gpt-audio-mini.
- Feb 23, 2026: GPT-Realtime-1.5 and gpt-audio-1.5.
- May 7, 2026: GPT-Realtime-2, Realtime Translate and Realtime Whisper.
- May 12, 2026: the Realtime API Beta removed.
- Jul 6, 2026: GPT-Realtime-2.1 and 2.1 mini.
- Sep 10, 2026: GPT-Live 1 generally available.
- Oct 16, 2026: prompt-to-voice and nine new GPT-Live voices.
For rate limits by tier, see API Rate Limits; for every price, OpenAI API Pricing.
Frequently asked questions
How much does the OpenAI Realtime API cost? As of Nov 6, 2026, gpt-realtime-2.1 costs $32 per 1M audio input tokens and $64 per 1M audio output tokens, gpt-realtime-2.1-mini $10 and $20, and GPT-Live 1 sessions $0.05 a minute billed per second, with backend model use extra.
What is GPT-Live in the OpenAI API? GPT-Live 1 is OpenAI's full-duplex voice model, generally available since Sep 10, 2026, that listens and speaks at once and hands reasoning and tool use to a separate backend agent through v1/live/sessions.
Does the OpenAI Realtime API use WebRTC or WebSocket? Both. OpenAI's Realtime guide uses WebRTC for browser and mobile clients and WebSocket for server connections, with an ephemeral client secret so the API key never reaches the browser.
Can the Realtime API translate speech? Yes. gpt-realtime-translate is a streaming speech-to-speech translation model, estimated at about $0.034 a minute as of Nov 6, 2026.
What happened to the Realtime API beta? OpenAI made the Realtime API generally available on Aug 28, 2025 and removed the Beta version on May 12, 2026, so older code that sent the beta header must move to the current interface.
Sources
- Realtime API, OpenAI, read Nov 6, 2026
- gpt-realtime-2.1 model page, OpenAI, read Nov 6, 2026
- gpt-live-1 model page, OpenAI, read Nov 6, 2026
- Models, OpenAI, read Nov 6, 2026
- Pricing, OpenAI, read Nov 6, 2026
- Changelog, OpenAI, read Nov 6, 2026
- Deprecations, OpenAI, read Nov 6, 2026
- Voice in ChatGPT, article 20001274, OpenAI Help Center, read Nov 6, 2026
- Deployment Safety Hub, OpenAI, read Nov 6, 2026
- GPT-Live-1 is now in the API, OpenAI, Sep 11, 2026 (video)
Moments
GPT-Live-1 is now in the API
OpenAI
Related
-
Batch APIThe OpenAI Batch API runs up to 50,000 requests from one file within 24 hours at 50% off the normal price. Flex processing gives the same discount on ordinary calls that can be slow, Fast mode costs double for quicker answers, and prompt caching cuts the price of repeated prompt prefixes.
- Fine-TuningOpenAI is winding down fine-tuning: the platform is closed to new users, new job creation was restricted on Jul 2, 2026, and existing customers lose job creation on Jan 6, 2027. Fine-tuned models keep serving until their base models are retired, and no GPT-6 model can be fine-tuned.
-
OpenAI APIThe OpenAI API lets programs call OpenAI's models (GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, image, voice and transcription models) with an API key, billed per token. New projects use the Responses API through the official SDKs, with documentation at developers.openai.com.
-
Responses APIThe Responses API is OpenAI's recommended endpoint for new projects, combining model calls, built-in tools, structured outputs and conversation state in one request. Chat Completions is still supported, and the Assistants API it replaced shut down on Aug 26, 2026.
-
AGENTS.mdAGENTS.md is a plain Markdown file of instructions for AI coding agents, a "README for agents", that Codex reads before doing any work. Codex loads a global ~/.codex/AGENTS.md and then every AGENTS.md from the repository root down to the working folder, closest last, up to 32 KiB; /init writes a first one. The open format is now stewarded by the Agentic AI Foundation.
-
CodexCodex is OpenAI's coding agent: it reads a codebase, edits files, runs commands and tests, reviews pull requests and works in the cloud. As of Nov 6, 2026 it lives inside the ChatGPT desktop app, in the Codex CLI, in IDE extensions, in Codex Cloud and on mobile, and every ChatGPT plan includes some of it, Free and Go too.
-
-
See also
-
-
- API Rate LimitsOpenAI API rate limits depend on your usage tier (Free, Build, Launch or Grow), which rises automatically as total credit purchases reach $5, $100 and $500. On the Build tier GPT-6 Astra allows 5,000 requests and 1,000,000 tokens a minute, and going over returns HTTP 429.
-
OpenAI API PricingOpenAI API prices are per 1M tokens: GPT-6 Astra costs $10 input and $50 output, GPT-6.1 Sol $2 and $10, GPT-6.1 Terra $1 and $5, and GPT-6 Luna $0.10 and $0.50, as of Nov 6, 2026. Batch and Flex halve those prices, Fast doubles them.
-
-
Apps SDKThe Apps SDK was OpenAI's 2025 toolkit for building apps inside ChatGPT on the Model Context Protocol (MCP). Since Jul 9, 2026 apps live inside plugins, and the Apps SDK documentation now opens OpenAI's Plugins docs, where you build an MCP server, optional UI and skills and submit them to the ChatGPT Directory.