Skip to content

ModelsxAI

xAI introduces Grok 3 Beta with Think reasoning mode and DeepSearch

xAI Grok 3 at a glance

Grok 3 and Grok 3 mini arrived in beta for X Premium subscribers, trained on the Colossus cluster, with a visible reasoning mode and a DeepSearch research agent.

What changed with xAI Grok 3

xAI's announcement post, dated February 19, 2025, introduced Grok 3 Beta and Grok 3 mini, both described as still in training. Reasoning variants are switched on with a Think button that lets users inspect the model's reasoning, and DeepSearch was presented as the first agent in the line, built to search and synthesize information.

xAI said Grok 3 was trained on its Colossus supercluster with about ten times the compute of previous leading models, and it reported a 1 million token context window. Rollout began with X Premium and Premium+ subscribers on X and Grok.com, with API access promised in the following weeks.

Where to compare xAI Grok 3

Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.

Why xAI Grok 3 matters for businesses

For businesses, Grok 3 added another frontier-class competitor, which helped push down prices and widen choice, though its launch figures were self-reported and the initial release targeted consumers rather than developers.

Sources for this story

  1. Official
  2. Grok 3 model page (opens in a new tab)

    Artificial Analysis · artificialanalysis.ai

    Benchmark

Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

  • #llm
  • #xai
  • #grok
  • #reasoning

Want this working in your business?

We turn new AI capabilities into agents, voice receptionists, and automations you can actually use.