xAI has been on a release spree: Grok 4.3 is now available to all API developers at $1.25/$2.50 per million tokens — an 83% cut from previous pricing — with native video input and document generation. Grok Build is a new coding agent CLI for SuperGrok Heavy subscribers. Grok Skills brings persistent, reusable AI workflows across conversations.
What Is Grok 4.3?
Grok 4.3 is xAI’s flagship model, now available to all API developers at dramatically reduced pricing: $1.25 per million input tokens and $2.50 per million output tokens — roughly 4x cheaper on input and 10x cheaper on output than competing frontier models. It features a 1-million-token context window, always-on reasoning, native function calling, and ranks #1 on the Artificial Analysis agentic tool-calling leaderboard.
What’s New in Grok Build?
Grok Build is a new coding agent and CLI for professional software engineering. Available in early beta for SuperGrok Heavy subscribers, it supports plan/approve workflows, clean diffs, MCP servers, and parallel subagents. It works with existing AGENTS.md files, plugins, hooks, and skills. A headless mode (-p) enables use inside scripts and automation pipelines.
How Do Grok Skills Work?
Grok Skills, announced May 18, let users define reusable instruction blocks — named capabilities with description and instruction fields — that Grok carries into every future conversation. Skills are importable as .zip, .skill, or .md files, making them shareable and version-controllable. This moves Grok beyond one-shot prompting into a programmable AI workspace with persistent expertise.
What Other Updates Came With Grok 4.3?
The model also supports native video input (up to 5 minutes in mp4, mov, or webm), document generation (PDF, PowerPoint, Excel, Word), and speech-to-text/text-to-speech APIs priced at $4.20 per million characters — 86-92% below OpenAI’s equivalent endpoints. xAI retired eight legacy models on May 15, consolidating around the 4.3 family.
Key Takeaways
- Grok 4.3 API at $1.25/$2.50 per million tokens — 83% price cut
- #1 on Artificial Analysis agentic tool-calling leaderboard
- Grok Build CLI with plan/approve workflows and parallel subagents
- Grok Skills enable persistent, reusable AI workflows across conversations
- Native video input, document generation, and speech APIs included
- Eight legacy models retired on May 15, 2026
Frequently Asked Questions
How does Grok Build compare to Claude Code or Codex? Grok Build supports similar workflows (plan/approve, diffs, MCP) but distinguishes itself with parallel subagents and worktree integrations. It’s currently in early beta for Heavy subscribers.
Can I use Grok Skills on mobile? Yes. Skills are available across web, iOS, and Android, synced to your account.
Is Grok 4.3 better than GPT-5.5? Benchmarks show Grok 4.3 leading on agentic tool-calling, while GPT-5.5 leads on graduate-level science and mathematics. The gap has narrowed significantly.