Elon Musk has announced the upcoming release of the beta version of their latest AI model, Grok 2. This announcement has generated considerable excitement within the AI community, particularly after Musk described the model as a “giant improvement” in response to a user’s question about its training data.
Despite the buzz, both Musk and xAI have kept details about Grok 2 under wraps. However, in March, Musk made a bold claim that Grok 2 would surpass current-generation AI models “on all metrics.” This statement has heightened anticipation for the model’s capabilities and performance.
The launch of Grok 2 is just the beginning, as it will be followed by Grok 3, which Musk has projected to release by the end of the year. According to Musk, xAI has been rigorously training Grok 3, a model that he believes will be on par with or even exceed the capabilities of GPT-5, the highly anticipated next-generation language model from OpenAI.
Grok 3’s development requires an immense computational effort, utilizing a vast cluster of 100,000 Nvidia H100 GPUs. Musk has expressed high expectations for the model, stating it will be “really something special.”
During a recent conversation on X Spaces, Musk revealed that as xAI scales up with Grok 3’s massive size, the company is encountering challenges related to data access. To address these issues, Musk mentioned that his team is exploring the addition of synthetic data or incorporating data from videos into Grok 3’s training corpus.
