# Andrej Karpathy Estimates 75% Barely Contact Large Language Models

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/30147
Published: 2026-10-12T09:11:08.000Z
Updated: 2026-10-12T09:11:08.000Z
Section: Technology

> His rough four-level estimate ranges from 1 billion to 2 billion casual users to about 5,000 people with internal access to frontier-grade systems.

Andrej Karpathy estimates that about 6 billion people, or roughly 75% of the global population, have barely come into contact with large language models, highlighting a wide gap in exposure to advanced AI tools.

The AI researcher described the figures as rough “napkin math” and organized them into a four-level funnel based on how people use AI and the systems they can access.

The broadest active tier includes about 1 billion to 2 billion people, or roughly 20% of the population. These casual users may use free chatbot products occasionally as improved search tools or writing assistants.

A professional tier includes roughly 20 million people, equivalent to about 0.2% of the population. In this group, AI agents can now complete large, complex projects that previously took weeks or months, with a prompt.

At the narrowest level, Karpathy estimates that approximately 5,000 people, or 0.00006% of the population, have internal access to frontier-grade systems. The estimate does not identify the users, companies or access standards involved.

Karpathy linked the differences in AI exposure to the size of the problems people give these systems, uneven capabilities across fields and access levels ranging from free to paid and internal systems.

“I have non-tech friends who (reasonably, imo) say they have not much use for it at all,” Karpathy wrote.

The estimates describe different levels of contact with AI capabilities rather than a measured count of global users. They also suggest that the most advanced applications remain concentrated among a small professional and internal-access group.