Prime Highlight
- xAI has officially launched Grok 4.1 across the Grok website, X, and mobile apps, bringing major improvements in creativity, emotional intelligence, and conversational quality.
- Grok 4.1 (Thinking) now leads the LMArena text-task rankings, surpassing models from Google, Anthropic, and OpenAI.
Key Facts
- Grok 4.1 significantly reduces hallucination rates, scoring 4.22% on real-world information questions and 2.97% on FactScore, far lower than Grok 4.0’s 12.09% and 9.89%.
- The model ranks first on the EQ Bench for emotional intelligence, with the standard version also taking second place.
Background
Elon Musk–led xAI has begun rolling out its latest AI model, Grok 4.1, after completing a silent launch across the Grok website, X, and mobile apps between early November and November 14. The company claims that the new version offers substantial improvements in creativity, emotional intelligence, and user-friendly interaction, resulting in smoother and more nuanced conversations.
Grok 4.1 has quickly made its mark on industry benchmarks. For the first time, it has overtaken both Gemini and ChatGPT on the LMArena leaderboard for text tasks. The “Thinking” version of Grok 4.1 holds the top spot, with the standard Grok 4.1 model coming in second. This places xAI ahead of strong competitors such as Anthropic’s Claude, Google’s Gemini, and OpenAI’s latest ChatGPT models.
The model also leads in emotional intelligence tests. Grok 4.1 (Thinking) took first place on the EQ Bench, and the regular Grok 4.1 came in second. Other models like Kimi K2, Gemini 2.5 Pro, and GPT-5 scored lower. In creative writing, Grok 4.1 took second and third place, while an early version of OpenAI’s GPT-5.1 won first place.
A key upgrade in Grok 4.1 is its reduced hallucination rate. xAI says the model scored 4.22% on real-world information questions, which is a big improvement compared to Grok 4.0’s 12.09%. On FactScore, which measures accuracy in biography responses, Grok 4.1 recorded a 2.97% accuracy rate compared to its predecessor’s 9.89%.
xAI says users will notice that the model feels “nicer, more helpful, and more understanding” in daily interactions. The launch comes shortly after OpenAI’s GPT-5.1 release, while Google is reportedly preparing Gemini 3.0.
Musk recently confirmed that xAI’s next major model, Grok 5, has been pushed to early 2026, after initially targeting the end of 2025.
