Elon Musk’s xAI has announced the arrival of Grok 4.1, the newest version of its AI model, and users are already noticing the difference. Musk shared the update on X, highlighting a major jump in speed,
stability, and answer quality. The company says this latest upgrade is designed to drastically cut down on hallucinations, one of the biggest challenges for modern AI systems. According to internal evaluations, Grok 4.1 is now not only smarter but also more dependable in creative, emotional and collaborative tasks. Here's everything you need to know about this.Grok 4.1 Becomes 3x Less Likely To HallucinateOne of the most talked-about improvements is Grok 4.1’s reduced tendency to present inaccurate information as fact. xAI says a big part of its development focus this time was on minimising hallucinations, especially for information-seeking prompts that demand accuracy. To do this, the team used real-world query data and ran the model through the FActScore benchmark, which includes 500 biography-based questions. The results were a clear leap forward. While the Grok 4 Fast model previously showed a 12 percent hallucination rate, Grok 4.1 brought that down to just 4 percent, a threefold improvement. The benchmark scores echoed this jump in reliability, with Grok 4.1 scoring 2.97 percent compared to the older model’s 9.89 percent.ALSO READ: Sundar Pichai, Jeff Bezos, Elon Musk Are Competing To Build Data Centres On MoonGrok 4.1 Tops Global AI BenchmarksThe company also tested the new model on LMArena, a respected benchmark used to compare top large language models. In the Text Arena tests, Grok 4.1 (in quasarflux mode) scored an Elo rating of 1483, placing it at the very top and outperforming all non-xAI models by a significant margin. Even its non-reasoning tensor mode secured the second position, ranking above the full-reasoning versions of competing models. Before public rollout, Grok 4.1 quietly went live between November 1 and 14, 2025, during which xAI monitored blind pairwise comparisons with its previous production model. In these tests, Grok 4.1’s responses were preferred 64.78 percent of the time, underscoring the upgrade’s impact.How To Access Grok 4.1The updated model is now available to all xAI users across grok.com, X, and the Grok mobile apps on iOS and Android. You can start using it simply by selecting “Grok 4.1” in the mode picker or choosing Auto mode, which will automatically load the most capable version.
/images/ppid_a911dc6a-image-176344705041467414.webp)

/images/ppid_a911dc6a-image-176344353259544938.webp)



/images/ppid_a911dc6a-image-176339103009375572.webp)
/images/ppid_a911dc6a-image-176335253574384223.webp)
/images/ppid_a911dc6a-image-176335955876143584.webp)
/images/ppid_a911dc6a-image-176336661498345894.webp)
/images/ppid_a911dc6a-image-176337356040461199.webp)
