Grok 3 review: is Elon Musk’s new AI model really better than GPT-4?
Grok 3 is officially here. Elon Musk’s AI model has already raised eyebrows with its ability to generate hyperrealistic images of famous people, including the CEO of X himself. Now, Grok has been upgraded with advanced reasoning capabilities, putting it in direct competition with the likes of OpenAI’s GPT-4.
During a livestream on X this Monday (Feb. 17), xAI introduced Grok 3, hyping it up as the best AI model out there. They claim it’s outperformed big names like OpenAI, Google, Anthropic, and DeepSeek on key benchmarks. And it looks like Grok 3 might actually talk the talk as it performed impressively under the codename “chocolate” in Chatbot Arena, a blind test where chatbots go head-to-head.
— xAI (@xai) February 18, 2025
Has Grok 3 launched?Musk says Grok 3 is still in beta, but users can expect upgrades literally every day. A voice interaction feature is reportedly just about a week away.
Subscribers to the X Premium+ plan, which was recently increased to $50 a month, were the first to get access to the model.
Is Grok 3 better than GPT-4?Grok 3 is said to be a huge leap from its predecessor, packing over ten times the computational power of Grok 2. It’s built to handle complex problems more effectively by breaking them down into smaller steps and double-checking its answers before responding.
Early tests show Grok 3 outperforming heavyweights like OpenAI’s GPT-4o, Google’s Gemini, and DeepSeek’s V3. It even comes with two unique reasoning modes: “Think,” which lets you see its thought process in real-time, and “Big Brain,” designed for tougher, more computation-heavy tasks.
On top of that, xAI has rolled out Deep Search, a next-gen AI search engine similar to what Perplexity, Gemini, and ChatGPT offer. And rumor has it, a synthesized voice feature for Grok is on the way soon.
To test the model, I asked OpenAI’s advanced reasoning model, o1, to come up with five prompts.
1. Logical reasoning and explanation
Grok 3 calculates the final distance as 23 miles with a step-by-step breakdown. Credit: xAI / ReadWrite
Prompt: “‘Two people start walking from the same point but in opposite directions—Person A walks at 3 mph, and Person B walks at 4 mph. After one hour, Person A’s speed increases to 5 mph, and Person B slows down to 3 mph. After 2 more hours, how far apart are they?’ Explain your reasoning step by step, showing exactly how you arrive at the answer.”
When presenting this puzzle to Grok 3, it stumbled almost immediately. The screen froze for a solid 30 seconds before coming up with a response. However, it did finally begin analyzing the data, correctly surmising that “the problem involves two distinct phases of walking: the first hour, and then two additional hours with updated speeds.” In the end, it managed to figure out the answer was 23 – the same as GPT-4’s response.
2. Contextual understanding and summarization
Grok 3 analyzes Sylvia’s emotional turmoil, offering a detailed summary and critique of the passage. Credit: xAI / ReadWrite
A structured analysis of Sylvia’s story, highlighting themes of trust and betrayal. Credit: OpenAI / ReadWrite
Prompt: “Read the following excerpt from a short story and write a concise summary that captures the main conflict and resolution. Then, critique the author’s writing style in one or two paragraphs.”
Grok 3 provided a fairly standard AI-type response for this prompt, using garden-variety language such as: “The author’s writing style is concise yet evocative.” However, it seemed to exceed GPT-4’s version by pointing out a glaring linguistic issue, stating: “The prose occasionally leans toward melodrama.” In this case, I think Grok 3 wins out.
3. Creative writing in a specific style
Grok 3 creates a whimsical cyber-fantasy tale set in the futuristic city of Neonspire. Credit: xAI / ReadWrite
A sci-fi fairy tale blending magic and technology in Neo-Aurelia. Credit: OpenAI / ReadWrite
Prompt: “Write a 200-word mini-story in the style of a whimsical fairy tale but set in a futuristic urban metropolis. Incorporate at least three imaginative elements that blend fantasy with advanced technology (e.g., holographic dragons, levitating forests, etc.). Aim for exactly around 200 words.”
Both Grok 3 and GPT-4 managed to produce a sci-fi tale that was under 200 words, and both were fairly average stories. Grok 3’s version was more adventure-driven, focusing on action and external goals, while GPT-4’s story is more reflective. Either way, none of these stories are likely to win a Pulitzer Prize, (lucky for us).
4. Real-time data analysis
Grok 3 generates a real-time traffic prediction report using AI-driven analysis. Credit: xAI / ReadWrite
AI models discuss the challenges of real-time traffic predictions, relying on simulated data. Credit: OpenAI / ReadWrite
Prompt: “Given real-time data streams from multiple sensors across a city (traffic, weather, and air quality sensors), predict the traffic conditions for the next 24 hours. Use historical data comparisons and current trends from the sensors to support your predictions. Present your findings in a detailed report.”
This is one area where Grok 3 surpasses OpenAI by a wide margin. For one, xAI has access to real-time information, allowing it to provide 15 separate sources to answer this question. On the other hand, whether it’s GPT-4 or GPT-4o, neither model can access real-time data and instead provides a simulation. Grok 3 wins this one, hands down.
5. Complex analysis
Grok 3 outlines a strategic energy transition plan focusing on policy, economics, and sustainability. Credit: xAI / ReadWrite
A roadmap for a country’s transition to renewable energy, balancing economic and environmental factors. Credit: OpenAI / ReadWrite
Prompt: “Examine the hypothetical case of a country transitioning from fossil fuels to renewable energy sources over a five-year period. Assume the country’s primary energy consumption is 50% coal, 30% natural gas, and 20% renewables at the start. Provide a high-level plan that outlines policy changes, economic considerations (like subsidies or job impact), and environmental goals. Conclude with potential challenges and how they might be addressed.”
Grok 3’s plans were much more specific in addressing the transition from fossil fuels to renewable energy sources. Not only did it calculate exactly how much governments would need to charge in carbon taxes and incentives, but it also provided a detailed breakdown of potential challenges, such as the possible loss of 100,000 jobs. In comparison, GPT-4’s response was far less impressive, relying mostly on hypotheticals.
Our verdictGrok 3 is chalking up to be a pretty formidable AI model, already outperforming in areas like access to real-time data—something GPT-4 lacks. That said, it still has fairly robotic responses to some of the more creative tasks. It’s still early days, but Grok 3 feels like it could be one of the big movers and shakers in the AI space, possibly disrupting things for OpenAI. Is it “scary good,” as Musk says? Not yet.
Is Grok 3 free?For a short time, Grok 3 is available for free to all! https://t.co/r5iLXi2pBm
— Elon Musk (@elonmusk) February 20, 2025
So far, we have been able to access Grok 3 beta mode without a premium plan. It appears that this is for a limited time only, however. Announcing the move on X, xAI posted “The world’s smartest AI, Grok 3, is now available for free (until our servers melt).”
At some point, users will need to subscribe to Super Grok to keep access. This premium tier gives early users a front-row seat to xAI’s latest AI updates and features. You can check it out through the Grok app or head over to grok.com to access it online.
Featured image: xAI / Canva
The post Grok 3 review: is Elon Musk’s new AI model really better than GPT-4? appeared first on ReadWrite.