Elon Musk’s AI company AI, Xai, has released its newest flagship AI model, Grok 3, late Monday night, along with new skills in the app grok for iOS and the site.
GrokXai’s response to models like Openai GPT-4O And Google Twinscan analyze images and respond to questions, and operate a number of functions on Musk’s social network, X. Grok 3, which developed for several months, was Optimistically planned For release in 2024, but missed that deadline.
Xai used a huge data center in Memphis – a data center containing about 200,000 GPUs – to train Grok 3. In a Post On X, Musk claimed that Grok 3 was developed with “10x” more computing than Grok 2, its predecessor, and with an enlarged training database that obviously includes recordings of court cases.
“Grok 3 is an order of magnitude more capable than Grok 2,” Musk said during a live-wing performance on Monday. “[It’s a] At most looking for AI, even if that truth sometimes opposes what is politically correct. ”
Grok 3 is a family of models, to be accurate – not just one. A smaller version of Grok 3, Grok 3 Mini, responds to questions faster at the expense of some precision. Not all models are available yet, but the rolling starts on Monday.
XAI claims that Grok 3 beats GPT-4O on references including Aime, which evaluates the performance of a model of sampling mathematical questions, and GPQA, which tests models with doctoral physical, biology and chemical questions. An early version of Grok 3 also won competitively at Chatbot Arena, a multiple test, which puts different AI models against each other and has users to vote on their favorite answers, according to Xai.
Two varieties of Grok 3, Grok 3 reasoning and Grok 3 mini -resonation, can carefully “consider” problems, similar to “reasoning” models like Openai O3 mini and Chinese AI company Deepseek’s R1. Reasoning models thoroughly bill themselves before giving results, which helps them Avoid some of the pits This normally travels models.
Xai claims that Grok 3 reasoning exceeds the best version of O3 mini-O3-mini high-on several popular references, including a newer mathematical reference called Aime 2025.
The reasoning models are accessible with the Grok app. Users can ask for Grok 3 “think” or – for more difficult questions – to exploit “Big Brain” mode for further, more careful reasoning. Xai describes the modes as best suited for mathematics, scientific and code-related questions.
Musk said that some of the reasoning processes are obscure to prevent distillation, a method used by AI model developers to extract knowledge from another model. Recently, Chinese AI Company Deepseek was charged with distillation of Openai models create their own.
Grok’s reasoning mode adheres to another new function called DeepSearch, Xai’s response to AI-powered “deep research” tools such as Openai’s deep research. DeepSearch scans the Internet and X to analyze information and deliver abstraction in response to a survey.
Subscribers to the Premium+ subscription will first acquire Grok 3, and other functions are inserted behind a subscription that Xai calls Supergrok. With a price of $ 30 per month or $ 300 a year, Supergrok unlocks additional reasoning and deep consultations and throws an unlimited image.
In the future – just after about a week from now – Grok will get a “vocal mode,” Musk said. A few weeks later, the GROK 3 models will arrive XAI corporate APIalong with the function of deepsearch. And a few months after that, Xai will open Source Grok 2, Musk said.
“Our overall approach is that we will open the latest version [of Grok]When the next version goes out, “Musk said.” When Grok 3 is ripe and stable, which is probably in a few months, then we will open source Grok 2. “
When Musk announced Grok two years ago, he launched the AI as rough, unfiltered and anti-“awakened”-generally, willing to answer controversial questions that other AI systems will not. He fulfilled some of that promise. Said to be vulgar, for example, Grok and Grok 2 would happily force, pushing a colorful language you probably wouldn’t hear Chatgpt.
But grok models before Grok 3 covered on political issues and will not cross some limits. In fact, one study found that Grok bent to the political left on issues such as transgender rights, diversity programs and inequality.
Musk blamed the behavior of Grok’s training data – public web pages – and promised “Change gross closer to politically neutral.” It is not yet clear if Xai has achieved that goal.
