Brainstorm in Silicon: Grok 4 Outsmarts Textbooks and PhD Minds

Elon Musk's xAI launched Grok 4, a cutting-edge AI model that's causing a worldwide buzz for its capacity to resolve engineering puzzles that even textbooks and search engines cannot comprehend.

Major Developments

Grok 4 is powered by Reinforcement Learning with Verifiable Rewards (RLVW), enabling it to learn by methodical trial and error

Musk claims Grok 4 operates at a PhD-level across disciplines and outperforms graduate students in reasoning tasks

The model scored 26.9% on Humanity's Last Exam unassisted, outperforming Google's Gemini 2.5 Pro and OpenAI's GPT-4. With assistance, Grok 4 Heavy scored 50.7%

Grok 4 demonstrated expertise in black hole collision simulation, sports game prediction, and video game creation

It integrates real-time data access and multimodal inputs, including text, images, and, in the future, video

Controversies and Criticism

Grok 4 has been faulted for echoing Musk's own views on divisive topics like geopolitics and immigration

They discovered questions that circumvented Grok's coherence with Musk's perspective, questioning bias and transparency

Earlier deployments of Grok were criticized for providing antisemitic responses, and xAI has since restricted social media presence and implemented increased security measures

Market Impact

Grok 4 is available on X Premium plans with two tiers at $30 and $300 per month

Its cooperative multi-agent architecture and performance criteria put it as a leading rival to OpenAI and Google

Sources: India Today, The Hans India, Indian Express, The Week, VentureBeat