xAI has officially launched Grok 4.1, the latest version of its large model. It is available for free on grok.com and the X platform. The company says the release represents a significant advancement in AI’s ability to think, respond, and engage with humans.
xAI has officially launched Grok 4.1, the latest version of its large model. It is available for free on grok.com and the X platform. The company says the release represents a significant advancement in AI’s ability to think, respond, and engage with humans.
Grok 4.1 topped LMArena's Text Leaderboard with a score of 1,483, the company said , making the new model significantly more effective at answering queries than any other chatbot on the market.
Grok 4.1 also came in first place in EQ-Bench3, a test of emotional intelligence that was evaluated by another AI, namely Claude Sonnet 3.7. This means that the model can respond in a more natural, empathetic and human way. As a result, users get conversations that are easier to understand.
A common problem with artificial intelligence is hallucinations, where the system provides incorrect or fabricated information. xAI claims that Grok 4.1 has reduced this problem by almost a third compared to the old version.
“In the process of further training Grok 4.1, we focused on reducing factual hallucinations in responses to information requests. Subsequently, we observed a significant reduction in the level of hallucinations for selected work-based information requests. To assess the level of hallucinations, we use a stratified sample of information requests from the work system. In addition, we evaluate the FActScore indicator, an open-ended test that contains 500 biographical questions about specific individuals,” xAI notes.
The model is currently available in automatic mode, meaning users don't need to do anything to start using it. Grok 4.1 can also be selected directly from the model selection menu.
