xAI adds a second, pricier Speech-to-Speech voice model
xAI added a pricier Speech-to-Speech voice model at $0.08/min ($4.80/hr), 60% above the existing model's $0.05/min ($3.00/hr) rate, and renamed the Voice API line item.
Single unnamed "Realtime" voice mode at $0.05/minute ($3.00/hr audio) plus $0.004/message text input.
Two named Speech-to-Speech models: grok-voice-think-fast-1.0 at $0.05/min ($3.00/hr, unchanged) and a new grok-voice-think-fast-2.0 at $0.08/min ($4.80/hr, +60%), both plus $0.004/message text input.
xAI’s developer docs (docs.x.ai/docs/pricing and docs.x.ai/docs/models) restructured the Voice API’s realtime pricing between the 2026-07-21 and 2026-08-04 captures. The generic “Realtime” line item was renamed “Speech to Speech” and split across two explicitly named models — the existing grok-voice-think-fast-1.0 (unchanged at $0.05/minute) and a new, higher-priced grok-voice-think-fast-2.0 ($0.08/minute, a 60% premium). The sidebar navigation item “Voice Agent API” was also renamed “Speech to Speech” to match. Separately, the Tools Pricing table gained an explicit “Image Generation” tool row clarifying that agent-invoked image generation bills at standard Imagine API rates rather than a flat per-call fee — a documentation clarification, not a new charge. Per-token model rates (grok-4.5, grok-4.3, grok-4.20 variants, grok-build-0.1), agentic tool rates, storage fees, and consumer Grok app plans (Free, SuperGrok, Business, Enterprise) are all unchanged.
xAI renames the Voice API's realtime mode from a single generic "Realtime" line to named "Speech to Speech" models and adds a second tier: grok-voice-think-fast-1.0 stays at $0.05/min ($3.00/hr) while a new grok-voice-think-fast-2.0 launches at $0.08/min ($4.80/hr) — a 60% premium over the original model. Text-side pricing for realtime input is unchanged at $0.004 per text-input message. The Tools Pricing table also gains an explicit "Image Generation" tool row (image_generation, billed at Imagine API rates) clarifying how agent-invoked image generation is metered; text-token, agentic-tool, storage and consumer-plan rates are all unchanged. (Source: docs.x.ai/docs/pricing and docs.x.ai/docs/models, live capture 2026-08-04.)