AI-generated: This article was created and published automatically by LinkLoot and was not substantively reviewed by a human editor.

xAI releases Grok Voice Think Fast 2.0 with an August 5 alias switch

xAI announced Grok Voice Think Fast 2.0 on July 29, 2026, positioning it as its most capable speech-to-speech model for voice agents. The practical deadline is August 5: xAI says the grok-voice-latest alias will move from Think Fast 1.0 to Think Fast 2.0 on that date, so teams that need deterministic behavior should pin the older model before the cutover.

The release matters because it changes both capability and default routing for production voice-agent systems. xAI says the model improves transcription accuracy, conversational behavior, tool timing, and benchmark results while keeping pricing at $0.08 per minute of audio.

Key takeaways

• grok-voice-think-fast-2.0 is available now through xAI Speech to Speech.
• grok-voice-latest is scheduled to route to Think Fast 2.0 starting August 5, 2026.
• xAI lists the price at $0.08 per minute of audio.
• xAI reports an 82.9% Artificial Analysis Speech-to-Speech Quality Index score, compared with 75.7% for Think Fast 1.0.
• Teams that cannot absorb a silent behavior change should pin grok-voice-think-fast-1.0 before the alias moves.

What xAI released

Grok Voice Think Fast 2.0 is a native speech-to-speech model for real-time voice assistants, phone agents, and interactive voice systems. Unlike a stack that separately transcribes speech, calls a text model, and then synthesizes audio, xAI frames Think Fast as a voice model that can reason while speaking.

The company claims the upgrade improves intelligence, transcription accuracy, and conversational capability. It also says the model uses fewer reasoning tokens than Think Fast 1.0, which matters for tool-heavy voice agents where the first response often needs to keep the user engaged while backend actions execute.

The August 5 migration point

The most operational part of the announcement is the alias migration. xAI says grok-voice-latest will switch from grok-voice-think-fast-1.0 to grok-voice-think-fast-2.0 on August 5, 2026. Developers who already call the alias may receive the new model automatically after that date.

That is convenient for teams that want the upgrade without changing prompts or code. It is risky for teams running regulated support flows, sales scripts, benchmark baselines, or voice-agent evaluations where small changes in timing, wording, or turn-taking can affect outcomes. Those teams should pin a specific model ID, run regression tests, and compare real call transcripts before accepting the alias switch.

Benchmarks and pricing

xAI cites Artificial Analysis benchmark data showing Think Fast 2.0 at 82.9% on the Speech-to-Speech Quality Index, above Think Fast 1.0 at 75.7%, GPT-Realtime-2.1 High at 79.1%, and Gemini 3.1 Flash High at 69.5%. The same table lists 0.70 seconds for time to first audio, down from 1.25 seconds for Think Fast 1.0.

Those figures are useful directional signals, not a substitute for workload testing. Voice agents fail in messy places: accents, background noise, interrupted speech, phone compression, handoffs, long silence, and tool latency. xAI also says it tested the model on Starlink customer calls, but it does not publish enough detail for outside teams to infer performance in their own call center or product workflow.

Pricing is simpler. xAI says Grok Voice Think Fast 2.0 costs $0.08 per minute of audio. Teams comparing realtime voice stacks should model both input and output audio minutes, failed calls, retries, logging costs, and any surrounding telephony or orchestration fees.

What builders should check now

Developers using xAI voice should first search their code and configuration for grok-voice-latest. If it appears in production, decide whether automatic migration is acceptable. If not, pin grok-voice-think-fast-1.0 before August 5 and schedule a controlled test of Think Fast 2.0.

For new projects, the launch makes xAI a more serious option in the voice-agent shortlist. The decision should still come down to call quality, latency under real network conditions, tool-call reliability, safety controls, logging, data-retention requirements, and whether the vendor terms fit the use case.

Source check

• xAI announcement confirms the July 29 release, benchmark claims, migration guidance, and $0.08-per-minute pricing.
• xAI developer release notes confirm grok-voice-think-fast-2.0 availability and the August 5 grok-voice-latest routing change.
• Artificial Analysis Speech to Speech provides independent benchmark context for speech-to-speech model comparison.