Elon Musks Ai Company, Xai, released his latest flagship -KI model Grok 3 and unveiled new functions for the GROK iOS and web apps.
PitchXai’s answer to models like Openais GPT-4O and Google TwinsCan analyze images and answer questions and a number of functions in the social network of Musk, X. Grok 3, which has been in development for several months optimistic planned For publication in 2024, but this deadline missed.
Monday is an ambitious start.
Xai has used a huge data center in Memphis, which contains around 200,000 GPU for training Grok 3. In A post Musk claimed on X that Grok 3 had been developed with “10x” more computing power than its predecessor Grok 2 using an extended training data set that is allegedly included Submissions from court proceedings.

“Grok 3 is an order of magnitude that is more capable than Grok 2,” said Musk during a live streamed presentation on Monday. “(It is a) maximum the truth searching AI, even if this truth sometimes contradicts what is politically correct.”
Grok 3 is a models family to be precise. A smaller version of Grok 3, Grok 3 Mini, answers more quickly to questions at the expense of accuracy. Not all models and related features of Grok 3 are still available (some are in Beta), but they started on Monday.
Xai claims that Grok 3 beats GPT-4O on benchmarks, including Aime (in which the performance of a model is evaluated on a sample of mathematical questions) and GPQA (in which models using physics, biology and chemical problems at PHD level be evaluated). An early version of Grok 3 also achieved competitive Chatbot ArenaA crowdsourcing test that, according to Xai, is correct against each other and coordinate the users on their preferred answers.

Two models in the new GROK 3 family, Grok 3 -Argumentation and Grok 3 Mini argumentation, can carefully “think through” problems, similar to “argumenting” models such as OpenAis O3 mini and Chinese KI company Deepseeks R1. Argumentation models try to check yourself before spending the results of what helps you Avoid some of the pitfalls This usually stumbles models.
Xai claims that Grok 3 surpasses the best version of O3-Mini-O3-Mini-High on several popular benchmarks, including a newer math tank called Aime 2025.

This argumentation models can be accessed via the GROK app. Users can ask GROK 3 to “think” or – more difficult – to use the “Big Brain” mode to obtain reasoning that uses additional computing. Xai describes the argumentation models as best suited for mathematics, science and programming questions.
Musk said that some of the “thoughts” of the argumentation models were covered in the Grok app to prevent distillation, a method used by AI model developers to extract knowledge from other models. Deepseek was recently accused of having distilled the models of Openai his own work.
The Grok argumentation models underpin a new function in the GROK app called DeepSearch, Xai’s response to AI-powered research tools such as Openais deep research. DeepSearch scans the Internet and X to analyze information and to provide a summary in response to a question.
Subscribers too X’s Premium+ Tier (50 US dollars per month) first get access to Grok 3, and other functions are connected behind a new plan, the XAIS supergroc is called. Price of $ 30 per month or $ 300 per year (When leaks are to be believed), Supergrok unlock additional argument and deachsearch queries and accuses the unlimited generation of image.

In the future – as soon as a week – the GROK app will win a “voice mode”, Musk said, which will give GROK models a synthesized voice. A few weeks later, Grok 3 models will be available Xais Enterprise APITogether with the DeepSearch ability.
Xai plans, open source Grok 2 in the coming months, said Musk.
“Our general approach is that we will make the last version (from GROK) open source open source when the next version is complete,” he continued. “If GROK 3 is mature and stable, which is probably within a few months, we will be open source grok 2.”
When Musk announced GRK about two years ago, he provided the AI model as nervous, unsuccessful and anti-waxed-in general to answer controversial questions that other AI systems did not answer. He delivered part of this promise. GROK and GROK 2, for example, being vulgar would like to be happy to hire and spit out a colorful language that they would probably not hear about Chatt.
But Grok models in front of Grok 3 secured to political subjects and would not cross Certain limits. Actually, a study found that GROK leaned on topics such as transgender rights, diversity programs and inequality.
Musk has held the behavior of the training data of GROK – public websites – and pledged “Show closer to politically neutral”. It is not yet clear whether Xai has achieved this goal and what the consequences could have.