Led by billionaire entrepreneur Elon Musk, xAI has raised the bar in artificial intelligence with the launch of its next-generation AI model, Grok 4. Officially unveiled this week, Grok 4 has quickly attracted attention for outperforming leading competitors on multiple benchmarks and is positioned as a direct rival to OpenAI’s forthcoming GPT-5.
In the challenging “Humanity's Last Exam,” Grok 4 achieved a 25.4% score, surpassing Google’s Gemini 2.5 Pro (21.6%) and OpenAI’s o3 (21%). Furthermore, in the tough ARC-AGI-2 benchmark focused on visual pattern recognition, Grok reached 16.2%—almost double the score of Anthropic’s Claude Opus 4 model.
Expanding the Grok family, xAI introduced “Grok 4 Heavy,” a special version equipped with multi-agent capabilities, and a new premium “SuperGrok Heavy” subscription plan—the industry’s most expensive at $300 per month. Subscribers have early access to Grok 4 Heavy’s cutting edge features. Grok 4 Heavy can generate multiple AI agents to collaboratively analyze complex problems and find the best solutions using group deliberation. With the "tools" feature enabled, Grok 4 Heavy scored a remarkable 44.4% on Humanity’s Last Exam, outpacing Gemini 2.5 Pro (26.9%).
In addition to these models, xAI plans to launch new coding, multimodal agent, and video generation models in the coming months, with direct API access for developers. With these bold innovations, xAI and the Grok family are set to shake up the AI industry once again.