Elon Musk’s xAI Unveils Grok 3, Claims It Outperforms OpenAI and DeepSeek

Elon Musk’s artificial intelligence startup, xAI, has introduced its latest AI model, Grok 3, which it claims surpasses offerings from OpenAI and China’s DeepSeek in early standardized tests on math, science, and coding.

The launch underscores Musk’s ambitions to challenge leading AI players amid an escalating race for dominance in the generative AI space.

At a demonstration streamed on his social media platform, X (formerly Twitter), Musk described Grok 3 as a massive leap forward compared to its predecessor.

“We’re very excited to present Grok 3, which is, we think, an order of magnitude more capable than Grok 2 in a very short period of time,” Musk said.

In addition to unveiling Grok 3, the xAI team also announced “Deep Search,” a next-generation AI-powered search engine, signaling the company’s push into the broader AI ecosystem.

The rollout of Grok 3 will begin immediately for premium X subscribers, while a separate subscription plan will be available for those who wish to access the model through a dedicated web and app platform, xAI confirmed.

Musk had previously teased Grok 3 at the World Governments Summit in Dubai, describing it as “scary smart” and highlighting its powerful reasoning abilities. He went as far as suggesting that Grok 3 might be the last AI model to surpass its own capabilities, hinting at the rapid advancements in AI self-improvement.

“This might be the last time that an AI is better than Grok,” Musk said.

He further revealed that the model was trained on a large amount of synthetic data, allowing it to reflect upon its mistakes and achieve greater logical consistency—an ability that has been a major challenge for AI systems.

Early testing results from Chatbot Arena, a crowdsourced AI ranking platform, reportedly placed an early version of Grok 3 ahead of competing AI models, though xAI has not provided independent verification of these claims.

While unveiling the product, Musk acknowledged that the model remains a work in progress but assured rapid improvements in the coming months.

“We should emphasize that this is kind of a beta, meaning that you should expect some imperfections at first, but we will improve it rapidly, almost every day,” Musk said.

He also confirmed that voice assistance capabilities would be introduced at a later stage.

Grok 3 Enters an Intensely Competitive AI Market

The launch of Grok 3 positions xAI as a serious challenger in the highly competitive AI space, which is currently dominated by OpenAI, Google DeepMind, and Anthropic.

Musk, who co-founded OpenAI in 2015, has since split from the company’s leadership over disagreements regarding its direction. His criticism has only grown louder, particularly after OpenAI transitioned from a nonprofit to a for-profit entity.1

In a recent high-profile move, Musk and a group of investors offered $97.4 billion to acquire OpenAI’s nonprofit parent company, an offer that OpenAI rejected. The feud between Musk and OpenAI CEO Sam Altman has further fueled speculation about Musk’s ambitions to create a direct OpenAI competitor through xAI.

OpenAI, for its part, continues to push forward with its own advancements. Last week, Altman provided a major update on the company’s roadmap for its next-generation AI models, revealing that GPT-4.5, codenamed “Orion,” will be the final model before OpenAI moves towards a more unified AI experience with GPT-5.

Meanwhile, Chinese startup DeepSeek recently shocked the AI industry by unveiling an open-source model that it claims can rival OpenAI’s o1 model, despite being trained using less powerful, more energy-efficient chips.

DeepSeek’s rise comes against the backdrop of U.S. sanctions that have restricted access to cutting-edge Nvidia AI chips in China. Despite this, DeepSeek has demonstrated that AI breakthroughs are possible without access to the latest Western hardware, which has broadened the competitive landscape.

xAI Expands Its Computing Power to Train Grok 3

To maintain its edge, xAI has significantly expanded its AI training infrastructure. The company had previously announced a Colossus supercomputer, powered by 100,000 Nvidia GPUs. On Tuesday, it revealed that it had doubled the size of its GPU cluster to train Grok 3, cementing its position as one of the most hardware-rich AI companies in the world.

While xAI’s massive computing power gives it an advantage, DeepSeek’s ability to achieve competitive results with lower-end hardware has led some experts to question whether brute-force GPU scaling is the most effective long-term AI strategy.

Some AI experts who spoke to CNBC praised DeepSeek for proving that efficiency can rival raw power, while others remain skeptical about whether its cost-saving methods can truly compete with well-funded giants like OpenAI and xAI in the long run.

Will Grok 3 Give Musk the Edge in AI?

Despite Musk’s ambitious claims about Grok 3’s superiority, xAI still has several hurdles to overcome.

First, xAI lacks the ecosystem reach of OpenAI and Google, whose models are deeply integrated into widely used consumer and enterprise applications. Microsoft, OpenAI’s biggest backer, has embedded GPT models into products like Windows, Office 365, and Azure, giving it a massive distribution advantage.

Second, OpenAI continues to lead in fine-tuning AI models for real-world applications, particularly in business and creative industries. While xAI markets itself as developing AI with a “truth-seeking” approach, it’s not clear how Grok 3 will be adopted by businesses and everyday users.

Elon Musk’s artificial intelligence startup, xAI, has unveiled its Grok 3 chatbot, calling it an “order of magnitude more capable” than its predecessor and the “smartest AI on Earth.” Its features will be rolled out to premium users of X — who now face higher prices — from Tuesday. Musk claims that in early testing, Grok 3 outperformed a model from China’s DeepSeek, as well as OpenAI, which is looking to fend off a takeover bid from Musk.