Grok 4 is a frontier large language model developed by xAI, released on July 9, 2025. It was designed to serve as a reasoning-heavy assistant with real-time access to information via the X platform. Trained on the Colossus supercomputer cluster—which utilized over 200,000 GPUs—the model represents a significant increase in compute and capability over its predecessors. It is characterized by its "first-principles" reasoning engine, aimed at improving logical consistency and mathematical accuracy.
The model architecture is multimodal, supporting the processing of text, images, and vision. It introduces a dual-mode system that allows users to toggle between a "Thinking" mode, optimized for complex puzzles and PhD-level STEM research, and a high-velocity mode for faster interactions. Performance benchmarks indicate high proficiency in coding tasks, with specialized variants like Grok 4 Code tailored for advanced developer workflows.
Grok 4 features a 256,000-token context window, while the Grok 4 Fast variant supports up to 2,000,000 tokens. The model serves as the primary intelligence for xAI’s ecosystem, including integrations with Tesla vehicles and standalone mobile applications. While the base model remains proprietary, xAI provides access through a unified API that supports function calling and structured outputs.