Elon Musk’s xAI launches Grok 4.1 with improved EQ, sharper creative writing
The upgraded AI model, rolled out over two weeks, claims two-thirds reduction in factual errors and superior emotional intelligence after novel training approach
Elon Musk’s xAI has launched Grok 4.1, a next-generation artificial intelligence model that is reportedly superior to other industry models in terms of emotional intelligence and creative writing while reducing factual errors by almost two-thirds.
Grok 4.1 has been rolled out across grok.com, X, and Android and iOS apps. The upgraded Grok was rolled out over a two-week period from November 1, and almost 65 per cent of the users reportedly chose Grok 4.1 over its predecessor.
Superior emotional intelligence
Grok 4.1 seems to have upgraded its emotional modelling, scoring 1,586 on EQ-Bench3, an emotional intelligence assessment tool, a big leap from Grok 4’s score of 1,206.
Also Read: Google boss Sundar Pichai cautions against blindly trusting AI
xAI says the superior performance is because of a novel training approach using advanced AI systems as training evaluators during reinforcement learning, which enabled the updated model to “refine conversational nuance and empathy”.
Improved factual accuracy
The factual accuracy has improved substantially, with hallucination rates dropping sharply from 12.09 per cent to 4.22 per cent on real-world queries.
This seems to be the result of overhauling the reinforcement learning systems and tightening the way Grok manages reasoning, logic, and step-by-step tasks.
Also Read: Explained: What is Google’s new Project Suncatcher to build AI data centres in space?
Users do not need to explain further or correct it as much as before.
Sharper creativity
Grok 4.1 has earned a 1,708.6 Elo rating on LM Arena’s Text Arena in the creative writing benchmark, placing it among the top AI models in the world.
Writers will discover that this model better structures long essays, articles, scene outlines, or scripts with clearer logic. Researchers and students get summaries that are less error-prone and with better foundation in the source material that is provided.
Also Watch: Watch | Can India, Indians become global leaders in gen AI?
xAI emphasised that Grok 4.1 “better interprets subtle user intentions and maintains a distinctive voice throughout extended interactions”.
Easier to trust
The upgrades in Grok 4.1 makes it a model that is easier to trust.
People tend to use a model more often and for more meaningful tasks if it avoids hallucinations, stays consistent, handles complexity without wobbling, and recognises emotional context, and Grok 4.1 scores better on these points.
Also Read: Explained: What is Elon Musk’s Grokipedia and how does it differ from Wikipedia?
xAI seems to be positioning Grok 4.1 more as a reliable assistant for everyday work than the biggest or most powerful model.