Elon Musk’s XAI launches Model Grok 3 against the background of a narrow AI competition

Elon Musk’s XAI on Tuesday revealed its most artificial intelligence model, Grok 3, claiming that it can surpass Openai and China Deepeek’s suggestions based on early tests, which include standardized tests for mathematics, science and encoding.

“We are very excited to introduce Grok 3, which we think is a more capable order than Grok 2 for a very short period of time,” Musk said in a demonstration of Grok 3, which was passed on on his social media platform X.

The team also said it is launching a new product called “Deep Search”, which will act as a “next -generation search engine”.

Grok 3 will be implemented for Premium X subscribers later during the day, and will also be available through a separate subscription to the Web and App version of the model, the XAI team said.

Speaking at the World Government Summit in Dubai last week were called the model “Terrible smart,” with powerful possibilities for reasoning, claiming that he is superior to all other existing models in the internal tests of XAI.

“This may be the last time AI is better than Gock,” Musk said at the time, adding that he was trained in “very synthetic data” and is able to think about his mistakes to achieve a logical sequence.

The XAI team claims that an early iteration of Grok 3 received better estimates from existing competitors Chatbot ArenaA website that publishes different AI models against each other in blind tests.

By the end of the demonstration of the product, Musk said the company would continue to improve the model.

“We have to emphasize that this is something like beta, which means that at first you have to expect some imperfections, but we will improve it quickly, almost every day,” he said, adding that the voice help for the model will be released in Late time.

Musk, vowed Regarding the potential dangers of artificial intelligence, the XAI began in 2023, entering the AI ​​generative market, which includes Openai Chatgpt.

In September last year Openai launched its most modern model O1 that came with reasoning And he was able to solve relatively complex tasks for science, encoding and mathematics.

Musk, along with Sam Altman, helped to create Openai as a non -profit in 2015.

However in recent years the management of Musk and Openai have been a feudS Musk recently runs an investor group that presented a proposal To buy a non -profit parent for AI Startup for $ 97.4 billion – the offer Openai refused.

Last month, Chinese launched Deepseek shocked the AI ​​market when I launched a technical book that claims that one of its open source models is able to oppose O1 O1 performance, although it uses a cheaper, less energy process S

He reached feats in the face of the United States, limiting the lead AI chipmaker Nvidia From the sale of avant-garde graphic processors-used for training of AI models-to China.

XAI has “Colossus supercomputer,“For AI training, which said last year used a cluster of 100,000 NVIDIA Graphic processors for AI training. On Tuesday, the company revealed that it doubled the size of its GROK 3 graphic cluster.

While many AI and technology experts have told CNBC This Depepeek has increased the competition of AI, showing what can be done with less advanced technologies, others are more shocking in terms of its impact.