Elon Musk’s Grok 3 Challenges OpenAI, DeepSeek, and Gemini
- Elon Musk’s xAI has launched Grok 3 with 10x more computing power than its predecessor, using over 100,000 Nvidia H100 GPUs.
- In benchmarks, Grok 3 outperforms Gemini-2 Pro, DeepSeek V3, and OpenAI’s GPT-4o in math, science, and coding.
- It’s currently available to Premium Plus subscribers on X, with an Android version in pre-registration and a SuperGrok tier expected soon.
xAI, the Elon Musk-led AI company, has launched its latest Grok 3 chatbot. Musk has described the Grok 3 as the “Smartest AI on Earth.” It succeeded the Grok 2, which was launched in August of last year. During the launch, Musk showcased the capabilities of the latest chatbot, which included advanced reasoning, text-to-video conversion, and self-correction mechanisms. So, let’s take a closer look at Musk’s latest AI offering:
What’s New with the Grok 3?
According to the company, the biggest upgrade is in the raw power. The Grok 3 packs 10 times more computing power compared to its predecessor. Here are some of its key upgrades:
- More Computational Power: Grok 3 uses over 100,000 Nvidia H100 GPUs, which deliver more than 200 million GPU hours for training. This constitutes a tenfold increase in computational resources compared to Grok 2, allowing faster processing of large datasets and improved accuracy.
- Better Reasoning Abilities: The model includes advanced reasoning capabilities and introduces a new feature called DeepSearch, which allows it to explain its thought process while responding to queries. This is something similar to Perplexity’s DeepResearch tool, which was launched the other day.
- Creative Capabilities: While the Grok 3 is good with logic and reasoning, it does not lag behind when it comes to creativity. It has developed a new game by combining Tetris and Puyo Puyo.
- Continuous Improvement: Musk stated that the Grok 3’s pre-training was completed in January and is already improving daily. He further mentioned that users will see the improvement in just 24 hours.
Grok 3 vs Gemini-2 Pro, DeepSeek-V3, GPT-4o, and Others
xAI also compared the Grok 3 with leading AI models like Google’s Gemini 2 Pro, DeepSeek V3, and OpenAI’s GPT-4o. Where Grok 3 was leading the chart in all three categories, which included science, coding, and math.
Specifically, in Math (AIME’24), Grok 3 scores 52, significantly outperforming the competition. Similarly, in Science (GPQA), Grok 3 leads with a score of 75, while other models score around 65 or lower.
In Coding (LCB Oct-Feb), Grok 3 continues its top position with a score of 57, surpassing competitors like Gemini-2 Pro, DeepSeek-V3, and others that score in the range of 34 to 41. Grok-3 mini also consistently outperforms the competition, but it lags behind the more powerful Grok-3, indicating a trade-off between size and performance.
Grok 3: Availability and Access
The latest Grok model is currently exclusive to Premium Plus subscribers on the X. However, users can access it through the Grok app on iOS or via the newly launched website at grok.com. Besides, the Android version is also in its pre-registration stage.
However, it’s not all the company is also planning to introduce a new subscription tier called SuperGrok. It is expected to offer even more advanced features for dedicated users.
FAQs
Where can I access Grok 3?
Grok 3 is available to Premium Plus subscribers on the X and via the website grok.com. There’s also an iOS app and an Android version in pre-registration.
How much is Grok 3?
The X’s premium Plus Plan includes access to Grok 3, which costs $50 per month. According to a report from TechCrunch hours after the launch of Grok 3 X hiked the price of its top tier Premium Plus subscription plan by a big margin.
Is Grok 3 better than ChatGPT?
If we go by the benchmarking chart shown by the xAI during the launch, then Yes, the Grok 3 outperforms GPT-4o in math, science, and coding tasks.