Grok: A STRATEGIC STEP TO ENHANCE DIRECT INVESTMENT READINESS

xAI has unveiled its latest flagship model, Grok 4, a direct competitor to industry giants like OpenAI's ChatGPT and Google's Gemini.

- 

The new model offers unparalleled performance with real-time search integration, native tool use, and frontier-level intelligence across both academic and practical benchmarks.

A NEW ERA

The release of Grok 4 marks a significant milestone for xAI, a company dedicated to accelerating human scientific discovery through the use of Al. Grok 4 was unveiled via a live stream as "the most intelligent model in the world" by Musk, with claims of PhD-level proficiency across all academic subjects. Unlike its predecessor, Grok 3, which focused on scaling next-token prediction, Grok 4 leverages an unprecedented 200,000-GPU cluster named Colossus to enhance its reasoning capabilities through advanced reinforcement learning.

Grok 4's performance stems from a combination of scaled-up compute and refined reinforcement learning techniques. Unlike Grok 3, which relied heavily on pretraining, Grok 4 was trained with a 6x increase in compute efficiency, enabling it to handle complex reasoning tasks at scale. The model utilizes a 256,000-token context window in its API (128,000 in the app), enabling extensive input processing, although it falls short of competitors like Gemini 2.5 Pro's 1 million tokens. Its training data was expanded beyond math and coding to include diverse domains, ensuring broader applicability.

A key innovation is Grok 4's native tool use, which allows it to integrate tools like code interpreters and web search seamlessly.