Skip to content
CallMissed

Explore CallMissed

Back to all models
LLM Chatnewvoicerealtime
OpenAI — gpt-live-1 logo

gpt-live-1

by OpenAI · Released 2026

OpenAI GPT Live — one model listens, reasons, calls tools and speaks over a single streaming session. Audio and text only, billed per minute. Voice-agent only.

OpenAI — gpt-live-1 logo
LLM Chat

gpt-live-1

Powered by OpenAI · Realtime speech-to-speech

Context Window

—

Parameters

Not disclosed

Max Output

N/A

Category

LLM Chat

Overview

`gpt-live-1` is OpenAI's GPT Live speech-to-speech model. A single session handles listening, reasoning, function calling and speaking, so there is no separate STT, LLM and TTS hop to stitch together.

It takes audio and text in, and returns audio and text out. Image and video input are not supported. Audio is raw 24 kHz mono 16-bit PCM in both directions, and transcripts stream as timestamped fragments on the session timeline rather than as finished turns — the application groups them.

Pricing is per minute: $0.05 per minute, billed per second. OpenAI publishes no per-1M-token rate and no context window for this model, so there is nothing to convert to a token price. Backend model and tool tokens are billed separately.

Fourteen voices are available — alloy, ash, ballad, coral, echo, sage, shimmer, verse, marin, cedar, juniper, breeze, vale and ember — with `marin` as the default. juniper, breeze, vale and ember are new in GPT Live and are not available on gpt-realtime.

It is voice-agent only: it is not callable on /v1/chat/completions. Capacity is limited, so it ships as an opt-in selection rather than a default.

Pricing

MetricPrice
Price /min₹5.0000

1 credit = ₹1 = $0.01 USD. Transparent per-call pricing — you pay only for what you call, with no seat fees or minimums.

Key Highlights

  • $0.05 per minute, billed per second
  • 14 voices — 4 of them new (juniper, breeze, vale, ember)
  • Audio + text in and out, function calling
  • Single streaming session — no separate STT/TTS hop

Technical Details

  • Model id: gpt-live-1
  • Audio: raw 24 kHz mono 16-bit PCM, both directions
  • Voices: alloy, ash, ballad, coral, echo, sage, shimmer, verse, marin, cedar, juniper, breeze, vale, ember (default marin)
  • Knowledge cutoff: 2025-07-31
  • Function calling supported; structured outputs not supported
  • No published per-token rate or context window — priced per minute

Strengths

  • Per-minute pricing with no token maths
  • Widest realtime voice set
  • Single-session speech-to-speech

Limitations

  • Voice-agent only — not callable on chat completions
  • No image or video input
  • No structured outputs
  • Limited capacity — opt-in, never a default

Use Cases

Live voice agentsPhone support botsTool-driven voice assistants

API Example

// GPT Live handshake — sent once the session WebSocket is open.
// Wait for session.started before streaming audio.
{
  "type": "session.start",
  "event_id": "evt_1",
  "session": {
    "model": "gpt-live-1",
    "instructions": "You are a helpful support agent.",
    "audio": { "output": { "voice": "marin" } },
    "delegation": { "type": "client" }
  }
}

Endpoint: POST /v1/voice/sessions (WebRTC) · WSS /v2/voice/agent · Model ID: gpt-live-1

Try gpt-live-1 now

Get 1000 free API credits on signup. No credit card required.

Start free Book a DemoRead docs