Michelle ThompsonElon Musk’s xAI has introduced Grok 3, a model claimed to surpass competitors like Google’s Gemini, OpenAI’s GPT-4o, Anthropic’s Claude 3.5, and DeepSeek V3 in math, science, and coding benchmarks. However, its performance relative to OpenAI’s o3-mini and DeepSeek R1 remains unclear.
Unveiled via a livestream on Musk’s X platform, Grok 3 features advanced web searching with “deep search,” the ability to develop online games, and a “big brain” mode for tackling complex reasoning tasks. The model is now available to users of X’s $40/month “Premium+” plan and subscribers of Grok’s standalone app or website.
The release comes as AI competition intensifies, with major tech firms expanding data centers to enhance their models. DeepSeek’s recent R1 model has also challenged Silicon Valley’s AI strategies by demonstrating more efficient training methods. Musk noted that Grok 3 was developed using 10 times the computing power of its predecessor, Grok 2, following xAI’s data center launch in Tennessee last year.
Describing Grok 3 as a “beta” release, Musk acknowledged initial flaws but promised rapid improvements. He also previewed an upcoming “voice mode,” similar to ChatGPT’s natural conversation feature, expected to launch soon.
Beyond AI, Musk continues to expand his influence, including his role as a “special government employee” under President Trump. Recently, he led an investor group attempting to acquire OpenAI for $97 billion—an offer rejected by the company’s board. Musk has criticized OpenAI for straying from its nonprofit mission and has called for greater transparency in AI development.
Regarding xAI’s own open-source plans, Musk stated that Grok 2’s code would be released publicly once Grok 3 reaches a “mature and stable” stage in the coming months.
