Last updated:

Grok

xAI's model family, notable for tight integration with X (formerly Twitter) and one of the largest disclosed training clusters in the industry.

Model Facts

Latest model: Grok 4.7

Provider
xAI
Released
Sep 21, 2026
Context window
500K tokens
Weights
Closed

Benchmarks

71.0%

DeepSWE v1.1

xAI-reported

Source

37.6%

Terminal-Bench 4.0

xAI-reported

Source

Full Model List

ModelReleasedNote
Grok 4.7Sep 21, 2026current frontier model for coding, agents and knowledge work
Grok 4.6Aug 12, 2026scored 61 on the Artificial Analysis Intelligence Index
Grok 4.5Jul 8, 2026500K-token context
Grok 4.20Mar 10, 20261M-token context
Grok-3Feb 17, 2025major reasoning-focused jump
Grok-2Aug 13, 2024first version broadly available via the X app
Grok-1Nov 4, 2023first public Grok release
Grok 5–still in training; Musk says 4.8 and 4.9 come first
Official model page:
x.ai
Official documentation:
docs.x.ai
Other sources:
Grok models explained

What it is

Grok is xAI's model family, tightly integrated with X (formerly Twitter). The current model is Grok 4.7, released September 21, 2026 as xAI's frontier model for coding, agentic tasks and knowledge work, with a 500K-token context window. It is available on the xAI API, in Cursor and in Grok Build.

What's next

Grok 5 has not been released and has no committed date; Elon Musk has said versions 4.8 and 4.9 will arrive before it.

Where these numbers come from

The scores are xAI's own published figures for Grok 4.7: 71.0% on DeepSWE v1.1 (high effort) and 37.6% on Terminal-Bench 4.0, up from 65.2% and 20.3% for Grok 4.6. They are vendor-reported, not independent tests.

Notable traits

  • Real-time-oriented integration with X.
  • Shipped as frequent point releases (4.5 → 4.6 → 4.7 within about three months).
  • Closed weights — available through X, the Grok apps and the xAI API.