Grok Voice agents now cost $0.05 per minute to build

Voice agents get a concrete price

xAI has put a concrete meter on Grok Voice agents: $0.05 per minute. That matters because voice-agent stacks are often assembled from separate speech-to-text, language-model, and text-to-speech services, making latency, cost, and debugging harder to reason about. A single no-code builder tied to Grok Voice gives developers a clearer unit of deployment and a visible starting price.

“Voice Agent Builder: a no-code platform to create human-like voice agents with Grok Voice. Available today at $0.05 / min.”

The source tweet was posted by xAI on July 1, 2026 at 15:33:21 UTC, inside this crawl’s 48-hour cutoff. The account normally posts Grok model, API, and product updates, so this item is more than a promotional clip: it changes what developers can build from the xAI console. Follow-up posts in the same thread said typical voice stacks combine three APIs, while Voice Agent Builder is one interface for Grok Voice, and that beta accounts include a free phone number to get started.

The concrete number is the important part. At $0.05 per minute, a 10,000-minute month would imply $500 in usage before any surrounding workflow costs. That gives startups and internal tool teams a simple first estimate for call-center experiments, sales qualification bots, appointment flows, or support triage. It also makes comparison easier against multi-vendor voice stacks where speech recognition, reasoning, synthesis, phone numbers, and orchestration may each be billed separately.

The limits are still visible. The tweet does not provide latency percentiles, supported languages, telephony regions, data-retention terms, uptime guarantees, or whether the builder exposes enough control for regulated customer-service workflows. Those details will decide whether this is mainly a fast prototyping tool or a production-grade voice platform.

Next, watch whether xAI publishes real latency numbers and enterprise controls for Grok Voice. A low per-minute price can attract testing quickly, but voice agents succeed only if interruption handling, transfer to humans, compliance logging, and failure recovery work in live conversations.