xAI releases Grok 3 model with performance on par with "AI Big 5"

As competition in the field of artificial intelligence heats up, Elon Musk's xAI has dropped another bombshell with the official release of its newest product, the Grok 3 model. This highly anticipated AI model not only achieves significant performance improvements, but also signifies that xAI is among the top AI labs in the world, competing with giants such as OpenAI, Anthropic, Google, and Meta.

The Grok-3 series of models are divided into 2 versions, Grok-3 and Grok-3 mini, with no official release of parameters and other information. Both versions of the model support inference mode, in which the Reasoning version of Grok-3 mini has been trained, while Grok-3 Reasoning is currently a beta version and has not been officially released.

Grok 3 Performance Jumps to Top AI Models

According to official information from xAI, the Grok 3 model will first be available through the Grok chatbot on the X platform and the grok.com website; API access is expected to open in the coming weeks, with pricing yet to be announced.

Of note, xAI revealed that Grok 3 was trained on its "Colossus" cluster of 100,000 H100 GPUs. This is likely the first model in the industry to be trained on such a large cluster of H100 GPUs (or equivalent computing power). The impressive performance improvement shows that the "law of scale" in AI still holds true, and that continued investment in computing power is expected to lead to a quantum leap in modeling capabilities, which is a major benefit for computing infrastructure providers such as Nvidia.

Grok 3 Official Published Review

Grok 3 Artificial Open Comparison Review Results

xAI Promoted to "AI Big 5"?

This is by no means a small technology iteration. Less than two years old, xAI is already at the forefront of technology with the release of Grok 3. xAI's rapid rise has allowed it to join the AI ecosystem built by OpenAI, Anthropic, Google, and Meta. xAI's rapid rise has allowed it to join the AI ecosystem built by OpenAI, Anthropic, Google, and Meta. xAI is now recognized as one of the "AI Big Five" in the United States.

Of course, it should be emphasized that the above analysis is based on performance data published by xAI. Currently, the industry has not yet been able to independently evaluate the actual performance of the Grok 3 series models, and their true capabilities have yet to be further verified.

Non-inferential modeling still has market potential

Despite the current interest and leadership of inferential models in the intelligence space, Grok 3 remains one of the top non-inferential models available. Non-inference models still have irreplaceable advantages and market needs when dealing with workloads that are more sensitive to latency and cost.

Grok 3 Live Demo

In order to better demonstrate the capability of Grok-3, we also give some demo screenshots.Grok-3 supports inference mode as well as deep inference mode, which looks good from the screenshots.

Demonstration of the Deep Research model of Grok-3

Grok-3's Reasoning Model 1

Grok-3's Reasoning Model 2

Grok iOS App Launches Simultaneously, Upgrading the Mobile Experience

To further expand Grok's application scenarios, xAI has also launched a new standalone Grok iOS app. With this app, users can experience powerful AI features on their mobile devices, such as generating beautiful images. Users can also log in to the Grok iOS app through their X account to personalize content such as real-time news, sports events, and local information for a more convenient and personalized information service.

iOS users in the U.S. can now download the Grok app from the App Store to get a sneak peek at the latest AI technology.

Grok iOS App Download Link