Recently, Elon Musk's xAI company announced the release of Grok4.1, an upgrade to the existing Grok4 model. This update introduces two new models, Grok4.1 and Grok4.1Thinking, both of which are available for free, but with fewer restrictions for paying users.

image.png

According to xAI, the probability of "hallucination" when generating content with Grok4.1 has been reduced by three times compared to the previous version, making Grok4.1 one of the best releases from xAI to date. However, it remains unclear how Grok4.1 performs against its competitors, such as the newly released GPT5.1, which has improved in emotional intelligence and performance.

image.png

According to preliminary performance data provided by the Text Arena tool from LMArena, Grok4.1 (Thinking) and Grok4.1 performed well in various competitive tests. Text Arena is an open-source tool that allows users to compare different large language models through side-by-side, blind testing, and randomized testing.

In the latest benchmark tests, Grok4.1 (Thinking) ranked first with a score of 1510, while Grok4.1 ranked 19th with a score of 1437. According to feedback from the testing platform, this version scored more than 40 points higher than the previously released Grok4Fast two months ago.

Although the upgrade of Grok4.1 is worth noting, it may not be the strongest model of the year. Google is preparing to launch Gemini3.0, which is expected to be the most powerful model to date.

Introduction: https://data.x.ai/2025-11-17-grok-4-1-model-card.pdf

Key Points:  

🌟 Grok4.1 is released, including two models: Grok4.1 and Grok4.1Thinking, both of which can be used for free.  

📉 The new model reduces hallucinations during content generation by three times, improving the accuracy of content generation.  

🏆 Preliminary benchmark tests show that Grok4.1 (Thinking) achieved the highest score, performing better than the previous version.