xAI launches Grok 4 and grok-build-0.1 coding API

xAI announced Grok 4, positioning it as the most intelligent model globally with native tool use and integrated real-time search, available to SuperGrok and Premium+ subscribers and through the xAI API. A new SuperGrok Heavy tier unlocks Grok 4 Heavy, the most powerful variant. Alongside it, xAI released grok-build-0.1 in public beta — the same model powering the Grok Build CLI — explicitly tuned for agentic coding and usable in environments like Kilo Code, OpenCode and OpenClaw.

The pricing is the headline for developers: xAI's official account confirmed grok-build-0.1 at $1 per million input tokens and $2 per million output, calling it 'extremely cost effective, intelligent, and fast.' That undercuts most frontier coding models and lands amid an intensifying price war led by DeepSeek's V4 Pro cut and Anthropic's Opus 4.8 reductions.

Mechanically, the integrated real-time search is xAI's differentiator — Grok pulls live data through native tooling rather than relying on stale training cutoffs, a pitch aimed at developers building current-events-aware agents. The coding focus directly targets Claude Code and OpenAI Codex's agentic-coding lead.

Competitively, xAI is trying to convert Grok's consumer base and X integration into a developer platform. Community reaction to the underlying corporate consolidation under SpaceXAI was muted, seen as restructuring with an unclear roadmap. The skeptical read: 'most intelligent model globally' claims need independent benchmark validation, and xAI's coding tools are unproven against entrenched rivals. Watch real-world adoption of grok-build-0.1 in third-party CLIs and whether the aggressive pricing holds once the beta ends.