# Realtime voice API pricing compared: GPT-Live 1, Grok Voice 3.0, Gemini 3.8 Live

Canonical: https://theagenticweekly.com/guides/realtime-voice-api-pricing-compared · Last verified: 2026-10-27 · Published: 2026-10-27 · Author: Agentic Weekly (operated independently from Texas, US) · Versions covered: Vendor pricing pages read 2026-10-27; prices change, check the pages before you budget · Evidence: Reported (list prices read from vendor pages, not run by us); no latency or quality measured · Tools: 

Checked 2026-10-27 by Agentic Weekly's editorial process: vendor pages re-read.

> OpenAI lists GPT-Live 1 at $0.05 per minute of live session, billed per second, with backend model and tool usage billed separately. xAI lists Grok Voice 3.0 at $0.08 per minute. Google bills Gemini 3.8 Live per audio token, about $0.005 per minute in and $0.018 out. Pages read 2026-10-27 (America/Chicago); the three are not like-for-like.

## The three list prices

Table: Realtime voice API list prices, read from each vendor's page on 2026-10-27

| Model | Price as listed | How it is billed | Source |
|---|---|---|---|
| OpenAI GPT-Live 1 (`gpt-live-1`) | $0.05 per minute of live session | Billed per second, not rounded up to a whole minute; backend Responses calls (model and tools) billed separately at their normal pricing | [OpenAI model page](https://developers.openai.com/api/docs/models/gpt-live-1) |
| xAI Grok Voice 3.0 | $0.08 per minute ($4.80 per hour) | Speech-to-speech API; same price as Grok Voice 2.0, per xAI's Oct 14, 2026 announcement | [xAI announcement](https://x.ai/news/grok-voice-3) |
| Google Gemini 3.8 Live (`gemini-3.8-live`) | $3.00 per 1M audio input tokens, or $0.005 per minute; $12.00 per 1M audio output tokens, or $0.018 per minute | Token-based; text input $0.75 and text output $4.50 per 1M tokens, image/video input $1.00 per 1M or $0.002 per minute; Search grounding extra | [Google pricing](https://ai.google.dev/gemini-api/docs/pricing) |

Google's page lists the same Gemini 3.8 Live rates for Gemini 3.8 Live Extended Thinking and Gemini 3.1 Flash Live Preview. We did not find a per-minute price for GPT-Realtime 2.1 on the pages we read for this guide, so it is not compared here; see [GPT-Realtime 2.1](https://theagenticweekly.com/models/gpt-realtime-2-1).

## Why these prices are not like-for-like

- **What a minute means:** OpenAI and xAI price a minute of session. Google prices the audio tokens in and out, so cost depends on how much each side speaks, plus any text, image or tool tokens.
- **Backend cost:** OpenAI says backend model and tool usage is billed separately, so $0.05 per minute is not the whole bill for an agent that reasons or calls tools. We did not find a backend line item on xAI's announcement; check the xAI pricing page for your setup.
- **Limits:** OpenAI's page lists concurrent-session limits of 50 (Build), 300 (Launch) and 500 (Grow), with the Free tier not supported.
- **What is not shown here:** latency, accuracy and voice quality. Vendors publish their own benchmark tables, for example xAI's Grok Voice 3.0 post cites Artificial Analysis figures; we did not measure any.

## Illustrative cost: 1,000 minutes of conversation

Our arithmetic from the list prices above, not a measured bill. It ignores backend model and tool costs, free tiers, taxes and any volume deals.

Table: List-price arithmetic for 1,000 session minutes

| Model | Calculation | Result |
|---|---|---|
| GPT-Live 1 | 1,000 x $0.05 | $50.00 plus backend model and tool usage |
| Grok Voice 3.0 | 1,000 x $0.08 | $80.00 |
| Gemini 3.8 Live, upper bound for audio | 1,000 x ($0.005 + $0.018), if audio flows in both directions for every minute | Up to $23.00, plus any text, image, tool or grounding charges |

The Gemini figure is an upper bound for audio only, because a call rarely has audio in both directions for the whole minute and Google bills tokens. A cheaper line does not make the model cheaper to run once backend, tool and grounding costs are added.

## How to choose

This is our reading of the prices, not a test result.

- Estimate minutes per month, then add each vendor's backend and tool costs, since a flat session price can hide them.
- Check concurrency limits for your tier before launch.
- Run your own call recordings through each model to compare latency and task success. For a local-model example of how we measure voice round trips, see our [voice round-trip study](https://theagenticweekly.com/research/voice-round-trip-piper-faster-whisper).
- For product-level picks across voice AI tools, see [best voice AI](https://theagenticweekly.com/best/voice-ai).

## What we tested

Nothing for this guide. We did not place calls, measure latency or compare voice quality. Prices come from the three vendor pages on 2026-10-27.

## FAQ

### How much does OpenAI GPT-Live 1 cost per minute?

OpenAI lists $0.05 per minute of live session, billed per second and not rounded up. Backend model and tool usage is billed separately at its normal pricing, so the voice session price is not the full bill for an agent.

### How much does Grok Voice 3.0 cost?

xAI's October 14, 2026 announcement lists Grok Voice 3.0 in the Speech to Speech API at $0.08 per minute, or $4.80 per hour, the same price as Grok Voice 2.0.

### How is Gemini 3.8 Live priced?

Google lists $3.00 per million audio input tokens, or $0.005 per minute, and $12.00 per million audio output tokens, or $0.018 per minute, plus text, image, video and grounding charges. It bills tokens, not session minutes.

### Which realtime voice API costs the least?

It depends on how you count a minute. At list prices Gemini's audio rates look lowest and Grok Voice 3.0's per-minute rate is highest, but OpenAI bills backend usage separately and Google bills tokens. Compare using your own call profile.

### Have you tested these voice models?

No. This guide compares list prices read on 2026-10-27 and our arithmetic on them. We did not place calls or measure latency, accuracy or voice quality, so we make no claim about which model sounds or performs better.

## Sources

- [OpenAI: GPT-Live 1 model page](https://developers.openai.com/api/docs/models/gpt-live-1) (fetched 2026-10-27)
- [xAI: Introducing Grok Voice 3.0](https://x.ai/news/grok-voice-3) (fetched 2026-10-27)
- [Google AI for Developers: Gemini API pricing](https://ai.google.dev/gemini-api/docs/pricing) (fetched 2026-10-27)

## Related

- [GPT-Live 1](https://theagenticweekly.com/models/gpt-live-1)
- [Gemini 3.8 Live](https://theagenticweekly.com/models/gemini-3-8-live)
- [Grok Voice 3.0](https://theagenticweekly.com/models/grok-voice-3-0)
- [GPT-Realtime 2.1](https://theagenticweekly.com/models/gpt-realtime-2-1)
- [Best voice AI](https://theagenticweekly.com/best/voice-ai)
- [Voice round-trip study](https://theagenticweekly.com/research/voice-round-trip-piper-faster-whisper)

## Changelog

- 2026-10-27: First published from OpenAI's model page, xAI's Grok Voice 3.0 announcement and Google's Gemini API pricing page, read on this date. Nothing was run.

## How to cite

Agentic Weekly, "Realtime voice API pricing compared: GPT-Live 1, Grok Voice 3.0, Gemini 3.8 Live", last updated 2026-10-27, https://theagenticweekly.com/guides/realtime-voice-api-pricing-compared