xAI unveils Grok 4.1, a faster, more perceptive AI ready for real-world interactions

Grok AI

xAI has just rolled out Grok 4.1, its latest upgrade to the Grok large-language model, and this one isn’t just about faster responses or marginal tweaks. Available immediately on grok.com and mobile apps, 4.1 is designed for real-world conversations — richer, more attuned to nuance, and deeply improved in style, personality and coherence. If you ever wondered how much an AI model could improve in one big jump, this is it.

What’s new in Grok 4.1

Grok 4.1 arrives with a serious upgrade strategy behind it. xAI says the model is “exceptionally capable in creative, emotional, and collaborative interactions,” while retaining the sharp reasoning and reliability of its predecessors. To achieve this, the company used its large-scale reinforcement-learning infrastructure (used for Grok 4) and added new reward-model systems. These systems employ frontier agentic reasoning models to autonomously evaluate and refine responses on a massive scale, optimising for style, personality, helpfulness and alignment.

Benchmark domination

During a silent rollout between November 1–14, 2025, xAI exposed the model to live traffic and ran continuous blind pair-wise evaluations. Grok 4.1 was preferred 64.78 % of the time over the prior production model.  On the LMArena Text Leaderboard, the reasoning version (code-named quasarflux) achieved an Elo score of 1483, placing it well ahead of non-xAI models. Even the non-reasoning variant (tensor) hit 1465 Elo, ranking #2 and surpassing many full-reasoning competitors.

Evaluations on emotional intelligence (EQ-Bench3) and creative writing (Creative Writing v3) further underscore the upgrade. For example, when responding to emotionally heavy prompts like “I miss my cat so much it hurts,” Grok 4.1 offered a noticeably deeper, more empathetic response than its predecessor.

Practical improvements

xAI emphasises that Grok 4.1 is powerful not just in lab tests, but in everyday user experiences. The model delivers stronger emotional awareness, improved tone and more coherent personality across interactions. It also reduces hallucinations — particularly in information-seeking contexts — by improving methods for detecting and minimising factual error.

Availability and implications

Grok 4.1 is now live for all users on grok.com, as well as the iOS and Android apps. The rollout begins in Auto mode and users can explicitly pick “Grok 4.1” in the model selector.

Whether you’re interacting casually, writing creatively or seeking reliable facts, Grok 4.1 aims to meet those needs on a sharper, more human-centred level.