xAI recently launched the latest iteration of Grok, its frontier AI model. We took Grok 3, claimed to be the ‘world’s smartest AI,’ for a spin, and here’s what we found.

Elon Musk’s xAI recently launched its frontier AI model—Grok 3. Dubbed as “the world’s smartest AI,” Grok 3 is far more capable than its predecessors. According to the company, it brings together strong reasoning and extensive pretrained knowledge. 

The AI model has been trained on xAI’s proprietary Colossus supercomputer cluster that possesses over 100,000 Nvidia Hopper GPUs. According to the makers, Grok 3 has shown significant improvements in reasoning, coding, mathematics, general knowledge, and tasks that require it to follow instructions. xAI has refined the chatbot’s reasoning abilities through large-scale reinforcement learning, which allows it to think for about a few seconds to minutes, mulling over the prompt before responding. 

The latest AI model from xAI has shown some stellar results across academic and real-world user benchmarks. Now, Grok 3 is available for all to try. Below is a compilation of things we tried with Elon Musk’s ChatGPT rival. 

First impressions

Much like DeepSeek-R1 and OpenAI’s ChatGPT, the homepage of Grok 3 has the input right at the centre, and it displays options— Attach file, DeepSearch and Think on the left side, and AI model picker and Enter options on the right. From the start, it is evident that a reasoning model has been embedded into it. Users can easily switch between standard AI functions and the “reasoning mode”. Strangely, the interface is eerily similar to ChatGPT.

On the web interface, one can spot the temporary chat in the top right corner before the history tab and profile icon. Temporary chat is a feature that allows users to access a conversation mode where their chat history is not saved. In this mode, all the conversations will be auto-deleted from the system within a 30-day window.