The Grok Blog

Notes from the lab.

Updates on what we're building, what we're researching, and what we've learned trying to make AI a little more honest, a little more useful, and a lot more fun.

G4
Release · Featured

Introducing Grok 4: real-time reasoning, on demand.

Our biggest jump yet. Grok 4 is faster, sharper on long-context reasoning, and now ships with Heavy — a multi-agent mode for the hardest problems. Here's what changed, what's next, and how to try it today.

By the xAI team May 8, 2026 · 9 min read
VX
Engineering

How Voice Mode handles interruption without breaking your train of thought.

Hana K.Apr 22 · 7 min
∑
Research

What we learned scaling Heavy's parallel reasoning to 32 agents.

Theo M.Apr 14 · 12 min
↯
Product

Grok inside X: how the timeline becomes a thinking partner.

Priya R.Apr 02 · 5 min
◐
Aurora

Aurora 2: faster image generation with sharper text rendering.

Liu Y.Mar 28 · 6 min
{}
Developers

Tool use, simplified: declare what you want, Grok handles the orchestration.

Marcus L.Mar 15 · 8 min
𝕏
Company

Our take on responsible deployment — and why we publish our refusals.

Jia W.Mar 04 · 11 min
↺
Changelog

What shipped in April: 14 things, ranked from "neat" to "actually huge."

The teamMay 01 · 4 min
✦
Customer story

How a small newsroom uses Grok to triage 4,000 tips a week.

Ana D.Feb 22 · 9 min

Get our notes,
before they're polished.

A short, occasional email when we ship, learn, or change our mind about something interesting. No marketing fluff.