Grok 4.20: Meet the team of four agents

Grok 4.20 Beta rolls out a fresh approach with four specialized agents that team up to handle queries more effectively. This multi-agent system boosts accuracy and depth in responses.

Grok acts as the team leader and final integrator. It coordinates the group, picks the right tools and strategies, pulls together inputs from the others, checks everything for quality, and puts out the polished final answer.

Harper serves as the research and analysis collaborator. It dives into web sources, X posts, and documents to collect solid information and deliver it in clear, organized ways.

Benjamin takes on verification and strategy, checking facts, spotting potential issues, considering different viewpoints, and ensuring the reasoning stays sharp and focused. Lucas handles logic and creative tasks and breaks down problems step by step, tackles technical details, and comes up with innovative ideas.

This multi-agent system changes how AI tackles complex questions. Instead of one model doing everything, these agents divide the work, much like a small team brainstorming.

For instance, when summarizing daily news, one agent might search for headlines, another verifies key events, a third analyzes regional impacts, and the leader weaves it all into a coherent report. Users see the process unfold in real time, with agents “thinking” through steps quickly—often in seconds.

The four-agent setup is available for premium and free users, while heavy users can go up to 16 agents working on the same query.

xAI launched this Grok 4.20 beta, which is available on grok.com, Android, and iOS apps. Early users report faster processing and more thoughtful outputs, though some mention the writing style hasn’t shifted much yet. You can try this new model and experience its capabilities.