Jonathan Laa | nuphoto | Getty Images
Elon Musk’s Xai announced its latest AI model, the Grok 3, on Tuesday, and could surpass its offerings from Openai and China’s Deepseek based on early tests, including standardized tests on mathematics, science and coding He insisted.
“I’m very excited to present the Grok 3. I think this is an order of magnitude more capable than the Grok 2 in a very short period,” Musk said of the Grok 3 streamed on social media. He spoke at the demonstration. Platform X
The team also said they are launching a new product called “Deep Search.” This acts as a “next-generation search engine.”
According to the Xai team, the Grok 3 will be rolled out for Premium X subscribers from Stateside on Tuesday and will be accessible via separate subscriptions for the model’s web version and app version.
Speaking at the World Government Summit in Dubai last week, Musk is called the “Scary Smart” model with strong reasoning capabilities, and claims it is superior to all other existing models in Xai’s internal tests .
“This may be the last time AI is better than Grok,” Musk said at the time, and was trained with “a lot of synthetic data” to achieve logical consistency. He added that it could reflect the mistakes.
The Xai team claimed that the early iteration of GROK 3 was given better ratings than existing competitors. Chatbot ArenaA crowdsourced website that pits different AI models against each other in blind testing.
Towards the end of the product demo, Musk said the company will continue to improve its models.
“We need to emphasize that this is a kind of beta version. That means you should expect some flaws at first, but it improves quickly almost every day,” he says, adding that voice support for the model. added that it will be released in A. later.
Fierce competition
Last September, Openai launched its most advanced model, the O1. Reasoning ability And it was able to solve relatively complex science, coding and mathematics tasks.
Musk, along with Sam Altman, helped create Openai as a nonprofit in 2015.
However, in recent years, Musk and Open leadership have rebutted. Musk recently led a group of investors who submitted a proposal to buy nonprofit parents of AI startups for $97.4 billion.
Last month, Chinese startup Deepseek argued that one of the open source models could rival the performance of Openai’s O1 model, despite using cheaper, energy-intensive processes. It shocked the AI market when it released.
It achieved a feat in the face of the US limiting major AI chip manufacturers in the US nvidia From selling cutting-edge GPUs used to train AI models to China.
xai says “Colossus SuperComputer,“Last year, the company said it was using clusters of 100,000 advanced NVIDIA GPUs for AI training. On Tuesday, the company doubled the size of its GPU cluster for training on the GROK 3. has been revealed.
Many AI and tech experts have told CNBC that Deepseek is showing what it can do with more advanced technologies, while others are more skeptical about its impact .
