xAI's Grok Build 0.1 API Unifies Text, Code, Voice, Images, and Video

xAI unveiled Grok Build 0.1, a new API that gives developers access to its frontier multimodal models through a single unified endpoint. The API supports text, code, voice, images, and video — all handled by the same model family, with models like grok-4.3 available out of the box. According to the announcement, the models were trained on the world's largest supercluster featuring 150,000 GPUs, enabling the system to process over 300 million queries daily. xAI emphasizes performance with a median latency under 200ms and the ability to handle more than 1 million API calls per day per user. The API is designed for easy integration, shown in their Python SDK example where a chat client is created with just a few lines of code. Pricing is usage-based, with rate limits that scale automatically. For enterprise teams, xAI offers custom rate limits, dedicated onboarding, audit logging, and data residency options via monthly invoices.

Grok Build 0.1 enters a crowded market already dominated by multimodal APIs from OpenAI (GPT-4o) and Anthropic (Claude 3.5). However, xAI differentiates with aggressive performance benchmarks and claims of running on the largest known training cluster. The single API for all modalities simplifies development workflows — instead of integrating separate models for vision, voice, and text, developers can now maintain one pipeline. Early adopters can use it for everything from building real-time voice assistants to generating code with visual context, all while benefiting from lower latency claims. The API is available immediately at api.x.ai with a free tier for initial testing.