Kavita Iyer Elon Musk’s AI company, xAI, has unveiled Grok 4, its latest and most powerful artificial intelligence (AI) model yet, claiming it now leads the field in academic, reasoning, and coding capabilities. Musk didn’t hold back, calling it the “smartest AI in the world” during a livestream on X (formerly Twitter) late Wednesday night. In addition to Grok 4, xAI rolled out Grok 4 Heavy, a more advanced variant that uses multiple AI agents collaborating like a virtual “study group” to solve complex tasks. The release also comes with a new top-tier subscription plan: SuperGrok Heavy, which offers access to this powerful model for $300 a month. Benchmark Battles: Grok vs. the Competitors According to xAI, both Grok 4 and its enhanced counterpart, Grok 4 Heavy, have outperformed major rivals like Google’s Gemini 2.5 Pro and OpenAI’s o3-high across a variety of leading AI performance benchmarks: Humanity’s Last Exam (HLE): Grok 4 scored 2%, nearly double the next-best model, Claude Opus 4. MMLU (massive multitask language understanding): Grok 4 achieved 6% accuracy and an Intelligence Index score of 73, topping the charts. “Grok 4 is the first time, in my experience, that an AI has been able to solve difficult, real-world engineering questions where the answers cannot be found anywhere on the Internet or in books. And it will get much better,” Musk posted on X. In the STEM and coding domain, the model appears even stronger: Grok 4 Heavy scored 100% on the AIME, a tough high-school level math test, while Grok 4 achieved 98.8%. On GPQA, Grok 4 scored 87.5%, while the Grok 4 Heavy variant edged it out at 88.9%. For coders, xAI teased the upcoming Grok 4 Code, which is set to launch in August 2025, and already shows 72–75% accuracy on the SWE-bench. While comparing PhDs to Grok 4, Musk said that “Grok 4 is postgraduate, like PhD level, in everything. Better than PhD — no exceptions. Most PhDs would fail where Grok 4 would pass.” Although he admitted that it still struggles with common sense and hasn’t invented new technologies or discovered new physics – “yet” but that is just a matter of time. New Pricing The Grok 4 API pricing will be the same as its predecessor — $3 per million input tokens and $15 per million output tokens ($0.75 per 1M cached input tokens). However, the bigger surprise lies in xAI’s updated subscription line-up: Free Tier gives users limited access to Grok 3. SuperGrok Plan ($30/month) unlocks access to both Grok 3 and the new Grok 4. SuperGrok Heavy ($300/month) offers full access to Grok 4 Heavy, Grok 4, and Grok 3, plus early previews of upcoming features. Grok 4 To Compete with GPT-5? xAI’s push to dominate the AI space comes just ahead of OpenAI’s expected GPT-5 launch later this summer. While Grok 4’s strong performance is hard to ignore, the question remains: whether businesses and customers will overlook recent controversies and choose Musk’s platform.