This is Grok 3, the most powerful AI from Elon Musk

The long-awaited presentation of Grok 3, the most powerful artificial intelligence of xAI, a company of Elon Musk, took place last night in the United States (our early morning) with a live broadcast featuring three team members to showcase this new tool to the world: “The mission with xAI and Grok is to understand the universe; we want to understand the nature of the universe to know what is happening, where the aliens are, what the meaning of life is, how it all started, and how it will end,” said the magnate, clarifying that this quest for truth will be pursued even if “sometimes it may be politically incorrect.”

Elon Musk
“The mission of xAI and Grok is to understand the universe.

We want to answer the biggest questions: Where are the aliens? What’s the meaning of life? How does the universe end?

To do that, we must rigorously pursue truth” pic.twitter.com/rgDQStnE3v

— Tesla Owners Silicon Valley (@teslaownersSV) February 18, 2025

Musk also introduced the Grok 3 Mini version, for less complex tasks and faster responses, along with the Grok 3 Reasoning Beta and Grok 3 mini Reasoning, intelligent search engines that conduct deep internet research to provide detailed reports, with performance similar to Perplexity.

Unfortunately, only the Premium+ users of X who pay an additional 33 USD monthly, or 330 annually, will be able to test this new tool. It is still not available in the United Kingdom or countries in the European Union.

While Grok 3’s possibilities are still being analyzed, below we explain its main features and how it compares next to its main competitor, ChatGPT.

What is Grok 3

Grok 3 is a chatbot designed to offer precise and up-to-date answers, integrating real-time information from the X platform. In addition to seeking the truth of the universe, this new tool from xAI aims to compete with DeepSeek, the Chinese AI, and ChatGPT, led by Musk’s arch-enemy, Sam Altman.

According to the owner of X, Grok 3 is “terrifyingly intelligent, the smartest on Earth.” It was trained with a computational capacity ten times greater than its predecessor, Grok 2, utilizing approximately 200,000 GPUs (graphical processing units). This increase in power allows Grok 3 to offer more advanced reasoning and more precise answers.

To put this in context, models like GPT-4 from OpenAI have been trained with tens of thousands of GPUs, but 200,000 GPUs is a massive amount even by current standards. This suggests that xAI has invested enormous computational capacity in Grok 3, aiming to surpass its competitors.

What you can do with Grok 3

Basically, Grok 3 allows you to perform the same tasks as other advanced AIs. The difference lies in the precision of its answers, the scope of its reasoning, the depth of analysis, the sources it relies on, and the quality of its results, both in text and image. Here are some of the things you can do with Grok 3:

• Generate content: Grok 3 can create texts such as stories, poems, scripts, and more, adapting to the style and tone requested by the user.
• Translate languages: It offers precise translations between multiple languages, facilitating global communication.
• Analyzing documents and graphs: It can examine and summarize complex information from various sources, including documents and graphs.
• Generating images: Integrated with the Aurora image generator, Grok 3 can create photorealistic images from textual descriptions, allowing for greater creativity and personalization. This is something that Grok 2, the previous version of this AI, which is currently available to all X users (paying or not), could already do. Let’s remember that this tool has generated several controversies due to its level of creative freedom, which allows for the generation of very realistic images of famous characters, facilitating the creation of deep fakes.

Performance Test Results

In evaluations, Grok 3 has demonstrated outstanding performance in various areas. In the AIME test, designed to assess advanced mathematical skills, the reasoning skills of Grok 3 and Grok Mini far outperformed Gemini-2 Pro, DeepSeek-V3, Claude 3.5 Sonnet, and GPT-4o, showcasing superior ability in solving complex mathematical problems, science, and coding.

Moreover, in the GPQA test, which measures knowledge in areas such as physics, biology, and chemistry, Grok 3 scored higher than its competitors, highlighting its extensive command of the natural sciences.

In terms of Grok-3 Reasoning and Grok-3 Mini Reasoning, they possess a deep thinking similar to that introduced by DeepSeek, allowing observation of how the AI develops its reasoning before generating a response.

While the benchmarks suggest that the reasoning capacity of Grok 3 will be one of the most advanced on the market, surpassing models like DeepSeek R1 and GPT-3 Mini, this functionality will be partially restricted to prevent other companies from analyzing and replicating xAI’s algorithms in their own artificial intelligence models.

Grok 3 Surpasses Other LLA Models in Arena.ai

LMArena.ai is an open evaluation platform designed to compare and evaluate large language models (LLMs) through anonymous and random matchups with community voting.

According to the evaluations that users have carried out of Grok 3, this AI has vastly surpassed the competition, receiving over 8,000 votes. Consequently, Arena.ai congratulated Elon Musk’s team on a post on the X social media platform:

BREAKING: @xAI early version of Grok-3 (codename “chocolate”) is now #1 in Arena! 🏆

Grok-3 is:
– First-ever model to break 1400 score!
– #1 across all categories, a milestone that keeps getting harder to achieve

Huge congratulations to @xAI on this milestone! View thread 🧵… https://t.co/p8z8lccNd5 pic.twitter.com/hShGy8ZN1o

— lmarena.ai (formerly lmsys.org) (@lmarena_ai) February 18, 2025

Photo: X.