xAI releases Grok 4.1

xAI/Grok

xAI released Grok 4.1 after a two-week silent rollout in which users preferred it over the previous model in 65% of blind comparisons. It topped LMArena's text leaderboard at 1483 Elo and shows fewer hallucinations on factual questions.

• Ranks #1 on LMArena's Text Leaderboard, 31 points ahead of competitors; non-reasoning variant achieves #2 without thinking tokens.
• Trained with large-scale reinforcement learning and frontier agentic reasoning models as reward models for evaluation at scale.
• Enhanced emotional intelligence via EQ-Bench3; measurable improvements on FActScore for biography-based factual accuracy.