Grok-4

MoE architecture with sparse + dilated attention. 256K token context. Trained with 100x more compute than Grok-2 and 10x more RL compute than Grok-3. Native tool-calling, real-time search, and extended reasoning.

Grok-4 Fast variant offers 2M context at $0.20/$0.50 per million tokens (AA Intelligence Index v4.3: 18). AA Intelligence Index v4.3: 22. Proprietary.