SAN FRANCISCO — xAI has released Grok 5, the newest and most capable version of its flagship model, with the company pointing to large gains in multi-step reasoning, coding, and raw response speed as it begins rolling the upgrade across its apps and API.
The launch arrives just days after xAI brought a new supercomputing cluster online, and executives credited that added training horsepower for the jump in capability. Grok 5 is positioned as a direct answer to the fastest-moving models in the industry, with xAI arguing it now leads on the reasoning benchmarks that matter most for real work.
What Is New in Grok 5
xAI says Grok 5 handles long, complex problems with fewer missteps, holding context across lengthy documents and chains of logic that tripped up earlier versions. The company highlighted stronger performance on math, science, and software tasks, along with faster first-token responses that make the assistant feel more immediate in conversation.
Much of that improvement traces back to the compute buildout behind the model. The Colossus 2 supercomputer that xAI recently fired up to train its next Grok gave the team room to scale training data and model size together, a combination the company says is the clearest path to better reasoning.
A Broader Grok Ecosystem
Grok 5 does not launch in isolation. It plugs into an expanding family of products, from everyday chat to creative tools like the image-and-video system behind the recently debuted AI-made series on Grok Imagine. The same underlying model now powers assistants across the lineup, so gains in Grok 5 ripple outward to writing, search, and generation tools at once.
