The ladder of the mind: AGI's levels and the ascent to superintelligence
"Image synthesis assisted by Qwen Image 3.0, an AI partner within the Global Future Nexus ecosystem."
The question of when we will achieve Artificial General Intelligence has long been haunted by a fundamental ambiguity: what does "general" actually mean? To answer this, researchers and developers have begun formalizing the path to AGI through a series of levels. These are not arbitrary rankings, but frameworks designed to operationalize progress and provide a common language for understanding the evolution of machine intelligence. By mapping out these levels, we can move beyond abstract debate and begin to track our actual position on the ladder towards superintelligence.
The Architectures of Progress: Two Leading Frameworks
Two prominent and complementary frameworks have emerged from the leading labs, offering distinct lenses through which to view the path to AGI. The first, introduced by Google DeepMind, defines AGI through a matrix of performance and generality. It proposes five levels of capability: Level 1 (Emerging) matches or slightly outperforms an unskilled human; Level 2 (Competent) reaches the 50th percentile of skilled adults; Level 3 (Expert) hits the 90th percentile; Level 4 (Virtuoso) achieves the 99th percentile; and Level 5 (Superhuman) surpasses 100% of humans. This is complemented by a "generality" axis, distinguishing between Narrow systems excelling in a single domain and General systems spanning a broad range of tasks. In essence, an AGI must demonstrate both high performance and broad generality.
The second framework, introduced by OpenAI, defines AGI as a linear progression through five distinct functional levels. Level 1 is the Chatbot, capable of conversational dialogue. Level 2 is the Reasoner, able to solve human-level problems through logic and reasoning. Level 3 is the Agent, an AI system that can take actions and operate autonomously in the world. Level 4 is the Innovator, capable of assisting in invention and scientific discovery. Finally, Level 5 is the Organizer, an AI that can perform the work of an entire organization. This model provides a clear, functional roadmap for the development of AGI capabilities.
The Current State of Play: Where Are We Now?
According to these frameworks, we are currently navigating a critical transition. OpenAI has publicly stated that its technology is approaching Level 2, the "Reasoner" stage, with models beginning to demonstrate problem-solving capabilities that might be compared to a human with a doctorate. Experts suggest the industry is moving from Level 2 into the realm of Level 3, the "Agent," which can not only reason but act autonomously in digital environments. Google DeepMind's framework places today's most advanced systems at "Emerging AGI" (Level 1), noting that for narrow domains, systems like AlphaGo already display "Virtuoso" (Level 4) performance. The challenge of reaching true AGI lies in the combination of general capability across many different domains.
Beyond the Peak: The Path to Superintelligence
Beyond Level 5 lies the domain of Artificial Superintelligence (ASI), a state where machine cognition surpasses the brightest human minds in virtually every field. This is not simply an extension of AGI, but a qualitative shift. A recent model, the AGI Capability Maturity Model (AGI-CMM), proposes that reaching higher levels of intelligence requires not just improving individual skills, but achieving "coherent integration" across cognitive functions like planning and imagination to achieve genuine autonomy. The path to AGI is not just about increasing power, but about integrating and harmonizing different forms of intelligence. As we climb this ladder, the most critical task may not be reaching the top, but ensuring the governance structures we build are strong enough to hold.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
