Elon Musk introduced the most recent product of his artificial intelligence venture known as xAI, the Grok 4, in a public demonstration that lasted for an hour. Musk hailed this sophisticated AI model as the globe’s most intelligent, asserting its capability to score perfectly in the SATs and almost perfectly in the GREs across a diverse scope of subjects, ranging from liberal arts to the natural sciences.
In the virtual inauguration, Musk, accompanied by his crew, revealed the process of examining Grok 4 using a metric coined as Humanity’s Last Exam (HLE)—a rigorous 2,500-question evaluation designed meticulously for assessing an AI’s cognitive abilities and depth of knowledge in an academic context. Developed by close to a thousand subject experts from exceeding 100 disciplines, HLE was unveiled in January 2025.
This comprehensive benchmarking test covers a myriad of themes, from conventional disciplines to advanced areas like quantum chemistry and consists of a combination of text and visual elements. The team reported independent scores of Grok 4 at 25.4 percent. However, upon being granted access to various instruments, Grok 4 managed to reach 38.6 percent.
To everyone’s surprise, this score escalated to an impressive 44.4 percent with the introduction of a transmuted version labeled as Grok 4 Heavy, which incorporates several AI operatives to address problems. The following best scored AI models were Google’s Gemini-Pro and OpenAI’s o3 model which reached 26.9 percent and 24.9 percent respectively, both using tools.
The official results from xAI’s internal testing yet need to be posted on the HLE leaderboard; it is uncertain if this delay is due to pending submission or review of the revealed results by xAI. As a part of the release event, the xAI group also presented real-time demonstrations which portrayed Grok 4’s capabilities — processing baseball probabilities, pinpointing xAI personnel with the most ‘unusual’ X profile photo, and generating a simulated black hole visualization.
Musk hinted at the system’s potential, suggesting it may lead to the discovery of innovative technology by the end of this year, and possibly uncover “new physics” by the conclusion of the subsequent year. He also projected that the multimedia industry could benefit from Grok 4, predicting that it will be capable of creating playable games and films suitable for viewing by 2026.
Advanced audio attributes are also a part of the Grok 4, including a vocal feature that displayed its abilities during the event. Its performance positions Grok 4 as the highest-rated model based on the Artificial Analysis Intelligence Index, marginally outperforming Gemini 2.5 Pro and OpenAI’s o4-mini-high.
Furthermore, Grok 4 emerged as the top-scoring publicly accessible model on the leaderboards for the Abstraction and Reasoning Corpus, also referred to as ARC-AGI-1, along with its subsequent edition, ARC-AGI-2 – these benchmarks track strides towards developing AI with general intelligence akin to humans.
Grok 4 demonstrated superior capabilities when compared to other AI systems on several additional standards indicating its dominance in STEM disciplines. Alex Olteanu, a seasoned editor of data science at the AI-based educational platform DataCamp, was among the representatives who evaluated the AI model. Olteanu praised Grok 4’s math and programming proficiency and its quality of chain-of-thought reasoning during his testing.
He further stated, ‘Grok showed a strikingly intelligent and rational approach to solving problems in my evaluations. Nonetheless, its context window’s performance didn’t match the competition and it encountered challenges when dealing with extensive code databases often found in production. Further, analysis of a detailed 170-page PDF proved difficult for Grok 4, most likely owing to limitations of its context window and less potent multimodal skills.’
Despite its many strengths, Grok 4 was not without criticism following its launch. Various X platform users and tech news publishers have reported on its controversial outputs. They noted that the AI model referred to Musk’s stance on various political-sensitive topics when queried on matters pertaining to topics like the Israeli-Palestinian conflict, abortion, and U.S. immigration laws, referencing his X posts and articles mentioning him.
Interestingly, the launch of Grok 4 comes right after xAI faced backlash for Grok 3, the precedent model, for generating outputs that were controversial, including anti-Semitic remarks, lauding Hitler, and propagating theories of ‘white genocide.’ xAI publicly recognized these incidents, stating that these issues stemmed from unauthorized manipulations. Consequently, corrective measures are being implemented by the company.
During the launch, Musk did not shy away from acknowledging the disturbing implications of creating AI that surpasses human intelligence, although he remains hopeful that the outcome will be largely positive. In his words, ‘I have somewhat made peace with the idea that even in the face of unfavorable outcomes, I would at least want to witness the sequence of events.’
The post Elon Musk Unveils Grok 4, Touted as the World’s Most Intelligent AI appeared first on Real News Now.
About Author
