Grok 4
xAI's frontier reasoner — Humanity's Last Exam leader at launch.
You're looking at an AI model record: who built it, what it can do, and the hardware and data behind it.
Grok 4 (July 2025) led hard-reasoning benchmarks including Humanity's Last Exam at release, with a Heavy multi-agent tier; Grok 4.1 (November 2025) tuned it for everyday use atop the X platform. Its rollout also featured high-profile alignment stumbles on X — a case study in deploying frontier models into a social feed. Details undisclosed. Point releases 4.3, 4.5 and 4.6 carried the line into mid-2026; Grok 5 — rumored at ~6T parameters, training on Colossus 2 — remained unreleased as of August 2026 with no committed date.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Developed by
Powers
Benchmarked on
Successor to
Sources
Record reviewed as of 2026-01-15 — see Methodology.