Affiliate disclosure: we may earn commissions when you sign up through some links below, at no extra cost to you. This never affects our pricing data, comparisons, or recommendations. Learn more.
news

xAI Grok Voice Think Fast 2.0 — 60% Price Jump (July 2026)

xAI launches Grok Voice Think Fast 2.0 with a 60% higher audio rate. The latest alias switches August 5, affecting unpinned apps.

By AI Pricing Guru Editorial Team

AI Pricing Guru articles are maintained by the editorial workflow behind the site: daily pricing snapshots, provider source checks, and review passes for model launches, subscription limits, and billing changes.

TL;DR

  • Grok Voice Think Fast 2.0 is available now through xAI's Speech to Speech API.
  • xAI lists 2.0 at a 60% higher audio rate than 1.0; the live table below pulls both prices from our dataset.
  • The grok-voice-latest alias moves to 2.0 on August 5, so unpinned production apps will inherit the new model and rate.
  • xAI's release note does not publish a quality or latency benchmark; canary the pinned 2.0 model before migrating.

Grok Voice Think Fast 1.0 vs 2.0 pricing

Model Audio / min Audio / hour Change
Grok Voice Think Fast 1.0 $0.05 $3.00 Baseline
Grok Voice Think Fast 2.0 $0.08 $4.80 60% higher

Minute-priced voice rows use the provider's published audio-minute rate. Additional text, tools, or telephony charges may apply.

Live values come from the canonical pricing API.

xAI has released Grok Voice Think Fast 2.0 for real-time speech-to-speech applications. The new model is available under the versioned ID grok-voice-think-fast-2.0; the moving grok-voice-latest alias will begin routing to it on August 5, 2026.

The migration has a direct pricing impact. xAI’s current voice pricing table lists Think Fast 2.0 at a 60% premium over the previous generation, so teams using the alias need to budget for a higher audio bill even if they make no code change.

What changed

Think Fast 2.0 runs through xAI’s bidirectional Speech to Speech API for voice assistants, phone agents, and interactive voice systems. Developers can pin the new model immediately or remain on grok-voice-think-fast-1.0.

The API supports WebSocket and WebRTC sessions, built-in or custom voices, server-side voice-activity detection, session resumption, configurable audio codecs, and tools including web search, X search, file search, MCP, and client functions.

xAI has not included a quality, task-success, or latency comparison in the release note. “2.0” is therefore an availability and pricing event today, not evidence that every voice workload improves enough to offset the premium.

Pricing comparison

The build-time table above reads both voice rates from our live pricing dataset. Think Fast 2.0’s listed audio rate is 60% higher than 1.0, while xAI lists the same separate text-input charge for both models.

xAI’s pricing page does not state a unit beside that text-input line, so teams should confirm how it appears on invoices before forecasting tool-heavy sessions. At any fixed volume, the audio portion rises by 60% before text input, tools, telephony, storage, taxes, or enterprise discounts.

Who benefits—and who pays more

Teams that need Grok’s native voice stack, X search, web search, remote MCP, or custom voices gain a new flagship route. New projects can test 2.0 without migrating an existing production session.

The immediate cost lands on apps that use grok-voice-latest. Once the alias changes, the same audio volume can cost 60% more unless xAI changes the published rate or the new model reduces conversation length, retries, or failed calls enough to compensate.

High-volume support and telephony teams face the largest exposure. A small difference per minute becomes material across long queues, abandoned calls, silence, hold time, and agents that fail to end a session.

What developers should do before August 5

  1. Pin grok-voice-think-fast-1.0 if predictable behavior and billing matter more than automatic upgrades.
  2. Canary grok-voice-think-fast-2.0 on production-shaped calls with the same voices, tools, VAD settings, and codecs.
  3. Measure resolved calls per dollar, not only price per minute. Track completion, transfers, interruptions, latency, retries, and average call length.
  4. Add maximum session duration, silence timeouts, escalation rules, and per-customer spend alerts.
  5. Confirm the text-input billing unit and tool charges on a small invoice before forecasting volume.

Use the xAI Grok pricing page for the wider model catalog. The AI token calculator covers text-model workloads, while voice sessions need separate minute, tool, and telephony estimates. For the competing premium voice stack, compare OpenAI API pricing and our OpenAI Realtime voice pricing analysis.

Labs availability

Think Fast 2.0 is not included in the current AI Pricing Guru Labs leaderboard. That suite scores deterministic text tasks through stable OpenRouter routes; it does not measure bidirectional audio, turn-taking, interruption handling, voice activity detection, or call resolution. xAI’s voice model also has no compatible text-completion route in the roster.

Publishing a proxy text score would misrepresent this launch. A valid future test needs a reproducible speech-to-speech harness with fixed audio prompts, end-to-end latency, interruption recovery, transcription accuracy, task completion, and cost per resolved call.

Alternative to benchmark

Teams evaluating conversational voice quality can benchmark ElevenLabs voice agents against the same call set before committing volume.

Affiliate disclosure: we may earn a commission from the sponsored link above. It does not affect the pricing analysis.

Bottom line

Grok Voice Think Fast 2.0 is available now, but the release note does not yet prove that it delivers a 60% improvement to match its 60% audio-price premium. Pin production workloads, run a controlled canary, and decide on cost per successful conversation before the latest alias moves on August 5.

Sources: xAI’s official release notes, Speech to Speech documentation, and voice pricing.