Grok Voice 3.0 Pricing: $0.08/min API Cost Breakdown
Kenji Tanaka
Inference Systems Writer

TLDR$0.08 per minute ($4.80/hour) plus $0.004 per text input, unchanged from Grok Voice 2.0. Cost math, rivals' per-task costs, limits, and open questions.
Grok Voice 3.0 Price Guide: $0.08 per Minute, $4.80 per Hour
Grok Voice 3.0 costs $0.08 per minute ($4.80 per hour) of speech-to-speech audio through the SpaceXSI (formerly xAI) Voice Agent API. Text inputs cost an extra $0.004 each. That is the same price as Grok Voice 2.0. The model launched on October 14, 2026 as grok-voice-3.0, and the grok-voice-latest alias now points to it. Prices on this page are current as of October 16, 2026.
Key Takeaways
- Headline rate: $0.08 per minute, or $4.80 per hour, confirmed in the official SpaceXSI announcement ("at the same price as its predecessor") and in the Artificial Analysis launch evaluation.
- Text surcharge: $0.004 per text input, according to developer writeups that quote xAI's pricing page.
- No price change: Grok Voice 2.0 users who move to Grok Voice 3.0 pay the same rate. The API endpoint, session parameters, and events are unchanged.
- Cost per Task: In Artificial Analysis's Big Bench Audio run, Grok Voice 3.0 costs $0.025 per task, compared with $0.030 for GPT-Live-1 (Astra, Medium) and $0.055 for GPT-Realtime-2.1 (High).
- Quality at that price: The model is #1 on the Artificial Analysis Speech-to-Speech Index with a score of 86.9, ahead of GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6.
Grok Voice 3.0 Pricing Breakdown
Grok Voice 3.0 is a speech-to-speech model billed by audio minute. It streams audio and text in both directions over a WebSocket, as described in the Speech to Speech docs.

Source: @0x0SojalSec
| Item | Price (as of Oct 16, 2026) | Source / note |
|---|---|---|
Speech-to-speech audio (grok-voice-3.0) | $0.08 / minute | SpaceXSI announcement |
| Hourly equivalent | $4.80 / hour | $0.08 × 60 min |
| Text input | $0.004 per text input | xAI pricing page, as quoted in developer writeups |
grok-voice-latest alias | $0.08 / minute | Points to Grok Voice 3.0 as of Oct 14, 2026 |
| Grok Voice 2.0 (deprecated) | $0.08 / minute | Retirement date not yet announced |
| Volume / enterprise discounts | Not yet announced | Custom rate limits are offered through Contact Sales |
| Per-voice or reasoning-effort price differences | Not yet announced | Reasoning defaults to "high" with no published price change |
Other Voice API products are billed differently. An August 13, 2026 snapshot of the xAI API page listed Text to Speech at $15.00 per 1M characters, Speech to Text at $0.10 per hour (REST) and $0.20 per hour (Streaming), and Speech to Speech "starting at $0.05 / min". That snapshot is older than the Grok Voice 3.0 launch. SpaceXSI has not explained how the $0.05 "starting at" figure relates to the $0.08 rate for Grok Voice 3.0.
What That Costs in Practice
All of the figures below are calculated from the published $0.08-per-minute rate and the reported $0.004 text-input charge. Call lengths and volumes are example assumptions. They are not vendor data.
| Scenario | Formula | Estimated cost |
|---|---|---|
| One 3-minute support call | 3 min × $0.08 | $0.24 |
| One hour of continuous conversation | 60 min × $0.08 | $4.80 |
| 10,000 minutes per month (small support queue) | 10,000 × $0.08 | $800 / month |
| One line open 24/7 for 30 days | 1,440 min × 30 × $0.08 | $3,456 / month |
| 20,000 calls per day at an assumed 3 min each | 60,000 min × $0.08 | $4,800 / day |
| 1,000 text inputs added to a session mix | 1,000 × $0.004 | $4.00 |
The 20,000-calls-per-day row uses the Starlink volume that SpaceXSI cites. The Verge quotes the announcement: Starlink call centers answer 20,000 calls a day with Grok Voice, and roughly 70% are resolved without a human. The 3-minute average call length is an assumption made for this page. SpaceXSI has not published call durations.
For most teams, the per-minute meter matters more than the text surcharge. At $0.004 each, 250 text inputs cost $1.00, which is the price of 12.5 audio minutes.
A community framing that circulated on launch day compares the $4.80 hourly rate with a cited ~$21.53 per hour median US customer-service wage (@AIRoboticsInt). That works out to roughly 4.5× cheaper per hour. The comparison leaves out integration work, escalation staff, and calls the model fails to resolve.
How Grok Voice 3.0 Pricing Compares
SpaceXSI says Grok Voice 3.0 "costs less than other frontier models." The data behind that claim is Artificial Analysis's cost-per-task measurement on Big Bench Audio, which SpaceXSI published in its announcement.
| Model | Cost per Task (Big Bench Audio, USD) | AA Speech-to-Speech Index | Time to first audio |
|---|---|---|---|
| Grok Voice 3.0 | $0.025 | 86.9 | 0.81 s |
| GPT-Live-1 (Astra, Medium) | $0.030 | 83.2 | 1.34 s |
| GPT-Realtime-2.1 (High) | $0.055 | Not yet confirmed | Not yet confirmed |
| Gemini 3.8 Live (Ext. Thinking, High) | Not yet confirmed | 82.6 | Not yet confirmed |
| Grok Voice 2.0 | Not yet confirmed | 81.3 | 0.70 s |
Cost-per-task and index figures come from the SpaceXSI announcement, which credits Artificial Analysis data dated October 13, 2026. The latency figures come from a launch summary by @0xLogicrw.
On Big Bench Audio, Grok Voice 3.0 completes a task for about 17% less than GPT-Live-1 ($0.025 vs $0.030) and about 55% less than GPT-Realtime-2.1 ($0.025 vs $0.055).
The cheaper model does not win on every benchmark. On τ-Voice, the support-call benchmark, GPT-Live-1 still leads with 74.5% against Grok Voice 3.0's 71.6%. SpaceXSI's response is that Grok Voice 3.0 reaches comparable τ-Voice results with 40% lower latency. Our earlier look at GPT Live 1's full-duplex design covers what OpenAI's side of that trade looks like.
One community dubbing test put Grok Voice at $0.012 against $0.111 for GPT Audio on the same script, about 9× cheaper (@thehypedotnews). That is a single unverified test, and the methodology was not published.
Free Tier, Limits, and Access
Where Grok Voice 3.0 runs: the Voice Agent API, the Voice Agent Builder, Grok Bot's voice mode, Grok in Tesla vehicles, and Starlink customer support, according to SpaceXSI. Developers connect at wss://api.x.ai/v1/realtime and pick the model with a query parameter. The docs show model=grok-voice-latest, and a versioned name such as grok-voice-3.0 pins a specific release.
Free access: The announcement page has a "Try It Free" button. No free API credit allowance for Grok Voice 3.0 has been published. Consumer-side pricing (Grok Bot or X subscription plans that include voice) is not tied to Grok Voice 3.0 in any official source reviewed for this page.
Rate limits: The xAI API page says self-serve accounts get "automatically increasing rate limits" and enterprise accounts get custom rate limits, invoiced billing, SSO, and data residency options. A third-party review reports two changes since Grok Voice 2.0: concurrency moved from a fixed 10 concurrent calls in one US region to a spend-based ladder that reaches 200, and an EU region was added. SpaceXSI has not confirmed this in the sources reviewed here.
Migration cost: The upgrade from Grok Voice 2.0 is free. The grok-voice-latest alias switched automatically, and the API surface did not change. Reasoning is on by default with effort set to "high", and it can be set to "none". No price difference between the two settings has been published.
Grok Voice 3.0 is speech-to-speech in a single model. Teams that run a separate text-to-speech stage for narration or outbound prompts can compare per-character pricing on a model like Gemini 3.8 Flash TTS, which is a different modality with a different billing unit.
What We Don't Know Yet
- Volume tiers: No discount schedule, committed-use pricing, or minimum spend has been published for $0.08 per minute.
- The $0.05 "starting at" figure: The pre-launch xAI API page listed Speech to Speech at "starting at $0.05 / min". It is unclear whether a cheaper configuration exists for Grok Voice 3.0.
- Per-voice pricing: Artificial Analysis benchmarked only the "Rex" voice. No source says whether cloned or custom voices cost more.
- Telephony costs: SIP phone calls are supported. Carrier charges are separate, and SpaceXSI has published no SIP surcharge.
- Cost per resolved call: Cost per task on Big Bench Audio is not the same as the cost of resolving a real support call. No public figure for cost per resolved call exists yet.
- Grok Voice 2.0 retirement date: The model is deprecated, but no shutdown date has been announced.
Frequently Asked Questions
How much does Grok Voice 3.0 cost per minute?
Grok Voice 3.0 costs $0.08 per minute of speech-to-speech audio through the Voice Agent API. Developer writeups quoting xAI's pricing page add a $0.004 charge per text input.
How much does the Grok Voice 3.0 API cost per hour?
Grok Voice 3.0 costs $4.80 per hour, which is $0.08 × 60 minutes. Text inputs are billed separately at $0.004 each.
Is Grok Voice 3.0 free?
No, not for API use. API access is billed at $0.08 per minute. The announcement page has a "Try It Free" button, and the model powers Grok Bot's voice mode, but no free API credit allowance has been published.
Is Grok Voice 3.0 cheaper than GPT-Live-1?
Yes, per task. In Artificial Analysis's Big Bench Audio run, Grok Voice 3.0 costs $0.025 per task, against $0.030 for GPT-Live-1 (Astra, Medium) and $0.055 for GPT-Realtime-2.1 (High).
Did Grok Voice 3.0 get more expensive than Grok Voice 2.0?
No. SpaceXSI released Grok Voice 3.0 at the same $0.08-per-minute price as Grok Voice 2.0, and grok-voice-latest users were moved over with no price change.
Are there extra charges beyond the per-minute rate?
Yes, text inputs are billed at $0.004 each, according to writeups quoting xAI's pricing page. Volume discounts, minimums, and per-voice price differences have not been published.
What to watch next
Three signals will change this page. The first is whether xAI's pricing page adds volume tiers or explains the $0.05 "starting at" Speech to Speech figure. The second is a retirement date for Grok Voice 2.0, which would set a deadline for pinned integrations. The third is OpenAI's API pricing for GPT-Live-1.1, which was announced on October 15, 2026 with API access "coming soon" and will reset the cost-per-task comparison. For broader context on the Grok family, see how Grok 4.7 compares with Claude Opus 5.
Try related models on Kie.ai: Gemini 3.8 Flash-Lite TTS and Gemini 3.8 Flash TTS.
About Kenji Tanaka
Kenji follows latency, throughput, and pricing signals to separate hype from shipped capability.
View all posts by Kenji Tanaka