Grok-3

model
xAI

Trained on 200K GPUs at the Colossus supercomputer with 10x more compute than Grok-2 and 12.8T training tokens. MoE architecture, estimated ~3T parameters. Up to 1M token context via API. Features Think mode (extended reasoning) and DeepSearch (agentic web research).

Grok-3 mini Reasoning achieved AA index 32. AA Intelligence Index: 25. Proprietary.

AnnouncementArtificial AnalysisOpenRouter

Model Details

Architecture MOE
Context window 1,000,000

Variants

Name Parameters Notes
Grok-3 — —
Grok-3 mini — Reasoning variant, AA index 32
frontierreasoningmoe

Related