Grok Voice 3.0 lands in the API at the same $0.08/min price

xAI's Grok Voice 3.0 speech-to-speech model is live in the Voice Agent API, tops Artificial Analysis's index, and trails GPT-Live-1 on τ-Voice.

xAI (now SpaceXSI, per SiliconANGLE) shipped Grok Voice 3.0 on October 14. It is available now in the Voice Agent API at $0.08/min ($4.80/hr), the same as Grok Voice 2.0. If you build support or booking voice agents, check one detail first. xAI says agents on the grok-voice-latest alias were switched over automatically, so your production model may already have changed.

Artificial Analysis says on X that the model debuts at #1 on its Speech to Speech Index at 86.9, ahead of GPT-Live-1 (Astra, medium) at 83.2 and Gemini 3.8 Live Extended Thinking (High) at 82.6. The same post gives two caveats. On τ-Voice, its simulated support-call test, Grok scores 71.6% against GPT-Live-1's 74.5%. Average time to first audio is 0.81s, faster than GPT-Live-1's 1.34s but slower than Grok Voice 2.0's 0.70s. H2S Media calls τ-Voice the test closest to the market xAI is selling into.

The cost picture is mixed. xAI's "cheaper" claim refers to Artificial Analysis's cost per task on Big Bench Audio. Artificial Analysis's own per-hour figures put Grok above Gemini 3.8 Live Extended Thinking and below GPT-Live-1 (Astra). TechRepublic points to two more costs in the API docs: a $0.004 charge per text input message, and billing for the full session length on default voice-activity-detection sessions. The Starlink results xAI cites (70% of calls resolved without a human) come only from the company.

What to do: pin a version if you need stable behavior. Then rerun your own support calls against GPT-Live-1, which Micdrop lists at $0.05/min plus its backend model, before you switch. Judge the models on resolved calls, not on the headline index.

Sources