Elon Musk’s xAI has just pulled back the curtain on its newest creation, Grok 4. The company boldly states this is currently the smartest AI model available. This puts it directly in the ring against OpenAI’s upcoming GPT-5 and Anthropic’s Claude 4 Opus. The AI landscape just got a lot more interesting.
A big talking point for Grok 4 is its improved accuracy. The xAI team put it through something called “Humanity’s Last Exam.” This test included tricky questions from math, science, and language. Grok 4 scored the highest among other leading AI models. What’s more, it did this without needing any extra tools to boost its answers. That’s a serious claim for raw intelligence.
Grok 4 comes in two main flavors. There’s the standard Grok 4. Then there’s “Heavy Mode,” which is designed for really tough problems. This mode can create several AI agents that work together to solve complex tasks. To access this powerful Heavy Mode, you’ll need the new “SuperGrok Heavy” package. It costs $300 per month. This is quite a jump from the current “Super Grok” Pro plan, which runs $30 per month. Grok 4 is available starting today. They also highlighted a Voice Mode, which responds well and sounds natural, even with a few limitations.
xAI isn’t stopping there. They’ve laid out a clear plan for Grok 4’s future, adding specialized abilities. A version focused on writing code will arrive in August. Following that, a model that can handle mixed types of data is set for September. Finally, a version capable of generating videos is expected in October. These focused updates show a clear strategy to dominate various AI tasks.