Google DeepMind proposes 6-level AGI framework to measure progress

DeepMind creates a roadmap toward artificial general intelligence

Google DeepMind published a new framework this week for measuring progress toward AGI (Artificial General Intelligence). The goal: establish a common standard for the entire industry, where five major AI labs currently use five different definitions.

Six levels — from zero to superhuman

The framework defines six AGI levels:

• Level 0: No AI
• Level 1 (Emerging): Equivalent to an inexperienced human
• Level 2 (Competent): Outperforms 50% of skilled adult professionals
• Level 3 (Expert): Top 10% of professionals
• Level 4 (Virtuoso): Top 1%
• Level 5 (Superhuman): Surpasses all humans

Each level is benchmarked across broad cognitive tasks — not just chess or narrow coding problems.

Why this matters

Without a shared measurement standard, it's impossible to compare progress, allocate investments wisely, or regulate the technology meaningfully. DeepMind believes the industry's lack of consensus on AGI definitions is a structural problem — and is trying to solve it.

For decision-makers: this framework could become the standard on which future regulatory rounds and investment decisions are based.

Source: WebProNews / Google DeepMind, March 2026