Elon Musk’s artificial intelligence startup, xAI, has recently unveiled its latest and most powerful iteration of the Large Language Model (LLM) called Grok-3. The announcement of Grok-3 was made during a livestream on the social media platform X.
Grok-3 is a significant advancement over its predecessor, Grok-2, as stated by Elon Musk himself during xAI’s livestream. The new LLM is a product of xAI’s Colossus supercluster, which boasts an impressive 200,000 GPUs, making it the world’s largest fully connected H100 supercluster of its kind. What sets xAI apart is its remarkable speed in scaling up its infrastructure, doubling the size of Colossus from 100,000 GPUs to 200,000 in just 92 days.
Pre-training for Grok-3 was completed in early January and involved over 10 times more training than Grok-2. In addition to the powerful LLM, xAI also introduced a smart search engine called DeepSearch with Grok-3.
Grok-3 is currently available to social media platform X’s Premium+ subscribers, with a new subscription tier called SuperGrok launched for the LLM’s mobile app and Grok.com website. Initial benchmark tests for Grok-3 have shown promising results, with the AI outperforming other large language models in its category. Notably, Grok-3 achieved a score of 1400 on Chatbot Arena, making it the first model to reach such a high rating.
Elon Musk has emphasized the importance of speed in innovation, highlighting xAI’s rapid progress despite starting several years behind industry leaders like OpenAI and Google. The company’s focus on accelerating development is evident in its achievements, with Grok-3 being just the latest example of xAI’s cutting-edge technology.
For those interested in staying updated on xAI and Grok-3, tips and news can be sent to simon@teslarati.com. Stay tuned for more exciting developments from Elon Musk’s groundbreaking artificial intelligence startup.