Grok 4.3 slashes prices, adds advanced voice‑cloning
- xAI rolled out Grok 4.3 and Custom Voices on April 30, pairing a new flagship model with instant voice cloning in its developer console. - The headline numbers are aggressive: Grok 4.3 is xAI’s recommended default model, while voice tools start at $3 an hour and TTS at $4.20. - This matters because model pricing is compressing fast, and xAI is bundling voice, cost tracking, and cloning into one API stack.
The real story here is not just “another model launch.” It’s that xAI is trying to turn Grok into a full application stack — text, voice, transcription, and cloned voices — while making the pricing hard to ignore. On April 30, xAI put Grok 4.3 forward as its default API model and launched Custom Voices, a voice-cloning tool built right into its console. A few days earlier and later, it also filled in the rest of the voice stack with speech-to-text, text-to-speech, and a faster voice agent. ### What actually launched? Two things landed together. Grok 4.3 became the model xAI says most API callers should use, and Custom Voices let users clone a voice from a short recording, then reuse that voice across xAI’s text-to-speech and realtime voice tools. That pairing matters because it turns Grok from “chat model” into “build a talking product here.” “cheap”? xAI’s public pricing page shows voice agent usage at $3.00 per hour, text-to-speech at $4.20 per 1 million characters, and speech-to-text at $0.10 per hour for REST or $0.20 per hour for streaming. The Grok 4.3 pricing lines are visible on the model page, but the scraped page excerpt doesn’t expose the token numbers cleanly, so the safest takeaway is the default, and the company is competing on cost as much as capability. ### What does “voice cloning” mean here? Basically, you upload a short reference clip and xAI creates a reusable voice ID. That cloned voice can then be called anywhere a built-in voice works — in plain TTS or in the realtime voice agent. xAI says clips can be up to 120 seconds, recommends 90 to 120 seconds for best results, and notes that the system copies not just timbre but delivery style too. ### Are there limits? Yes — and they matter. Custom Voices is currently available only in the United States, except Illinois. xAI also says custom voices are scoped to your team, not shared globally, and users can create up to 30 custom voices for free in the console. That suggests xAI wants fast adoption, but with some guardrails around rights there. ### Why bundle this now? Because the market is shifting from “best chatbot” to “best bundle.” A model alone is easier to swap out. A model plus realtime voice, cloning, transcription, and billing hooks is stickier. xAI also added per-request cost tracking on April 30, exposing exact request cost in API responses. That sounds small
