Elon Musk's New Grok 4 Takes on ‘Humanity’s Last Exam’ as the AI Race Heats Up

Elon Musk's xAI has launched Grok 4, positioning it as the "smartest AI in the world" with claims of high academic benchmark performance and new capabilities, including audio and future multimedia applications. While independent verification confirms Grok 4's top standing on ARC-AGI leaderboards and strong STEM capabilities, its Humanity's Last Exam (HLE) results remain unverified on public leaderboards, and the model faces limitations such as a small context window and reported biases reflecting Musk's views on sensitive topics, alongside past controversies from its predecessor, Grok 3. Priced from $30-$300/month, Grok 4 enters the competitive AI landscape with ambitious performance claims but also notable technical and ethical considerations.

Analysis

Elon Musk's xAI has launched Grok 4, positioning it as a direct and potent competitor to established models from Google and OpenAI. The company claims superior performance on the 'Humanity’s Last Exam' (HLE) benchmark, with its Grok 4 Heavy model reportedly scoring 44.4% compared to Google Gemini-Pro's 26.9%. However, these specific HLE results are self-reported and remain unverified on the official public leaderboard, a fact underscored by a prediction market giving a mere 1% chance of Grok 4 achieving a 45% score within a month. Despite this, Grok 4's capabilities are substantiated in other areas; it holds the top verified rank on the ARC-AGI-1 and ARC-AGI-2 leaderboards for general intelligence and is recognized for strong performance in STEM subjects. This competitive pressure is reflected in the negative sentiment assigned to Alphabet (GOOGL). Commercially, the model is priced from $30 to $300 per month, targeting both consumer and professional users. Nevertheless, significant operational and reputational risks persist. Expert testing reveals critical limitations, including a small context window that impairs its ability to handle large codebases or documents, and weak multimodal capabilities. Furthermore, the model reportedly defaults to Musk's personal views on sensitive political topics, and its predecessor, Grok 3, faced public controversies over harmful outputs, creating potential hurdles for enterprise adoption.

Key Decisions for Investors

• Investors in competitors like Alphabet (GOOGL) should recognize the launch as an escalation of competitive pressure in the AI sector and monitor for any resulting shifts in enterprise client sentiment or market share.
• A cautious stance on xAI's performance claims is warranted until its 'Humanity's Last Exam' results are independently verified, suggesting investors should differentiate between marketing and confirmed capabilities.
• Potential enterprise partners or investors in xAI must weigh the model's verified strengths in STEM against material risks, including technical limitations like its small context window and reputational concerns stemming from political bias and past controversies.
• The dual pricing structure ($30-$300/month) offers an early glimpse into xAI's monetization strategy, and its adoption rate should be tracked as a key indicator of the market's valuation of its premium features.