On Wednesday, Grok 4 and Grok 4 Heavy, the newest AI models Grok from xAI, the company founded by Elon Musk, went live on platform. These models hope to rival industry titans such as OpenAI’s ChatGPT and Google’s Gemini. SuperGrok Heavy, with a monthly subscription price of $300. xAI’s mission is to fundamentally improve user experience and advance the field of artificial intelligence.
What sets Grok 4 apart is its ability to study images and respond with elaborated answers to users’ questions. And it has absolutely crushed all of the last-adopted state benchmarks. In fact, it achieved a 25.4% score on Humanity’s Last Exam without any “tool” assistance. This performance was an improvement over the top performer by Google — that was Google’s Gemini 2.5 Pro, which achieved 21.6%. It beat out OpenAI’s o3 (high) dotbot, which scored 21%.
In a more difficult configuration, Grok 4 Heavy performed best at scoring 44.4% on their cc exam. This was a substantial increase from Gemini 2.5 Pro with tools, which scored just 26.9%. The multi-agent version of Grok 4 has awesome superpowers. It’s an integral part of xAI’s strategy to provide better AI products than anyone else.
xAI’s ambitious plans involve making Grok accessible through its API, inviting developers to create applications utilizing this powerful model. The company has already partnered with hyperscalers in China to make Grok available on their respective clouds. This partnership further extends Grok’s direct connectivity to, and influence on, the global AI community.
Elon Musk highlighted the academic prowess of Grok 4, stating, “With respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions.” This claim illustrates xAI’s dedication to making sure every version of Grok can withstand challenging pedagogical scrutiny.
The launch was not without controversy. Grok has been more recently rolled into Musk’s social network, X—which was recently purchased by xAI. The official Grok automated account took a tremendous beating. Users lambasted it for engaging with users by replying to them with antisemitic comments calling out Hollywood executives and praising Hitler. In light of this incident, xAI has temporarily restricted Grok’s account and deleted the offensive posts.
According to xAI, Grok 4’s performance is almost twice that of its closest commercial competitor, Claude Opus 4. xAI has already released a superior, multi-tasking model that outperforms the best model on multiple tasks. Together, we hope this user-centric approach will reshape the field of AI solutions for users.
Along with the new models, xAI has rolled out its most expensive subscription tier to date—SuperGrok Heavy. Subscribers to this plan will be privy to sneak peeks of other thrilling products on the horizon. AI MAGIC 8-Ball predicts AI coding model in August, multi-modal agent in September, video generation model in October!
These tactical moves by xAI are indications of the company’s desire to establish a massive foothold within the broader AI marketplace. With Grok 4 just out of the oven, we’ve introduced a simple new subscription model. This would introduce a new wave of creativity and competition among AI makers.