What you need to know

Elon Musk’s artificial intelligence startup, xAI, has launched the most powerful iteration of its Large Language Model (LLM), Grok-3.

Grok-3’s was formally announced in a livestream on social media platform X.

Grok-3:

As per Elon Musk during xAI’s livestream, “Grok 3 is an order of magnitude more capable than Grok 2.”
Grok-3 is the product of xAI’s Colossus supercluster, which is currently comprised of 200,000 GPUs.
xAI noted that Colossus is the world’s largest fully connected H100 supercluster of its kind.
What is quite amazing is that xAI doubled the size of Colossus from 100,000 GPUs to 200,000 in just 92 days.
For context, it took xAI 122 days to get the first 100k GPUs running for its Colossus supercluster.
Grok-3 completed pre-training in early January. It involved over 10x more training than its predecessor, Grok-2.
xAI also introduced a smart search engine with Grok-3, called DeepSearch.

https://t.co/hEfQ31gANQ

— xAI (@xai) February 18, 2025

Grok-3 is currently being offered to social media platform X’s Premium+ subscribers.
xAI has also launched a new subscription, SuperGrok, for the LLM’s mobile app and Grok.com website.
Initial benchmark tests for Grok-3 have been very optimistic, with the new AI outperforming other large language models in its category.
Grok-3 also reached a score of 1400 on Chatbot Arena, the first model to achieve such a rating.

During the xAI Engineering Open House event in San Francisco last year, Elon Musk shared a quote that stood out to the event’s attendees.
During his talk, Musk reportedly emphasized that speed in innovation is of the utmost importance.
“No SR-71 Blackbird was ever shot down and it only had one strategy: to accelerate,” Musk wrote.
Ultimately, it is difficult not to be impressed with xAI’s speed. After all, it has only been 17 months since the company started kicking off its very first model.
xAI also started its operations several years behind industry leaders such as OpenAI and Google.