The company xAI has introduced a new version of its AI model Grok 4.1, which is now available to all users for free on grok.com , X, and in mobile apps. Grok 4.1 was released on November 17, 2025, after a two-week stealth launch, during which it already outperformed the previous version in 65% of user preference tests. The model topped the LMArena Text Leaderboard, and its “quasarflux” mode scored 1483 Elo points, which is 31 points higher than any non-xAI model.
Grok 4.1 received record scores for emotional intelligence — 1586 on the EQ-Bench3 test, and 1722 on Creative Writing v3, which is 600 points higher than the previous xAI result. Developers reduced the number of factual errors to a third of the previous version, making this model the least prone to incorrect answers among all Grok models.
Users who have already switched to Grok 4.1 note noticeably faster responses and better quality, with greater empathy and creativity in role-playing tasks, writing, and daily queries. The model demonstrates an increased level of emotional intelligence, empathy, and interpersonal skills, as confirmed by its high scores in specialized tests.
xAI created Grok 4.1 based on the large reinforcement learning system used in the previous version, but focused on personality consistency, more accurate intent recognition, and autonomous reward modeling. As a result, Grok 4.1 has become the new standard in blind human preference tests.
All users, regardless of account type, can select Grok 4.1 in the model selection menu or allow the app to activate it automatically. This continues xAI’s policy of making the most capable models available to the widest possible audience without paid subscriptions.
