The short answer
Right now, “superintelligence” has four live meanings, and they are routinely conflated in coverage:
- The research concept — a system that greatly exceeds human cognitive performance in virtually all domains of interest. A research concept; this leaderboard does not establish its achievement.
- The U.S. government term — since September 29, 2026, the official executive-branch name for today's AI as a whole, per Executive Order 14434.
- Corporate lane-names — Meta Superintelligence Labs, Microsoft's MAI Superintelligence Team, Safe Superintelligence Inc.: directions of travel, not claims of arrival.
- The colloquial sense — simply “an extremely high degree of intelligence” (Merriam-Webster's second sense).
The research concept (Bostrom, 2014)
The term entered the modern AI discussion through Nick Bostrom's 2014 book Superintelligence: Paths, Dangers, Strategies (Oxford University Press), which defines a superintelligence as an intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest. That framing — systems beyond human capability across the board, with open questions about control and safety — remains the standard in AI-safety and research writing. See Bostrom's own summary and Wikipedia's article.
On the record as of Oct 8, 2026: the cited Nature editorial frames it as a goal, not an established reality:
“superintelligence is an aspirational category of systems, not a reality.” — Nature, “The science of superintelligence and why AI is still a useful word” (2026)
The corporate lane-names
Since 2024, major labs have attached the word to their frontier efforts. These are statements of direction — each dated and linked:
- Safe Superintelligence Inc. (founded June 2024) — Ilya Sutskever's lab made the
term its name and entire product:
“Superintelligence is within reach. Building safe superintelligence (SSI) is the most important technical problem of our time.” — ssi.inc (retrieved Oct 8, 2026)
- Meta (June 2025) — announced Meta Superintelligence Labs, and keeps a standing
product page:
“Over the last few months we have begun to see glimpses of our AI systems improving themselves. The improvement is slow for now, but undeniable. Developing superintelligence is now in sight.” — meta.com/superintelligence (retrieved Oct 8, 2026)
- OpenAI's Sam Altman has used the word for years, in the one-word spelling:
“It is possible that we will have superintelligence in a few thousand days (!); it may take longer, but I'm confident we'll get there.” — “The Intelligence Age”
- Microsoft (November 2025) — formed the MAI Superintelligence Team, and drew its
own line past AGI:
“If AGI is often seen as the point at which an AI can match human performance at all tasks, then superintelligence is when it can go far beyond that performance.” — Mustafa Suleyman, “Towards Humanist Superintelligence”, Nov 6, 2025
The U.S. government term
On September 29, 2026, Executive Order 14434 made “Super Intelligence” (two words) and “SI” the official executive-branch vocabulary for AI generally — defined not as a future capability but as the technologies covered by the existing statutory definition of artificial intelligence. That is a different, much broader meaning than Bostrom's, and it now appears on NIST, NSF and White House pages. Full quotes and analysis on our policy explainer, and every dated adoption step on the statements tracker.
Frequently asked questions
Is superintelligence the same as AGI?
Different authors draw the line differently. In the research sense, AGI usually means matching human ability across most tasks, while superintelligence means exceeding it across virtually all domains of interest. Microsoft's 2025 essay puts it this way: if AGI is matching human performance at all tasks, superintelligence is going far beyond it. The two words are often used interchangeably in casual coverage, which is part of the confusion.
Has superintelligence been achieved?
This leaderboard does not establish that research-sense superintelligence has been achieved. Nature's 2026 editorial on the terminology dispute calls superintelligence an aspirational category of systems, not a reality. In the new U.S. government sense, “Super Intelligence” is simply the official name for today's AI systems as a whole — so under that definition the term already applies to current models.
Why are there two spellings — “superintelligence” and “Super Intelligence”?
Researchers and companies write “superintelligence” (one word): Bostrom's book, Safe Superintelligence Inc., Meta's product page. The U.S. executive branch writes “Super Intelligence” and “SI” (two words) under Executive Order 14434 of September 29, 2026. This site uses one word for the research concept and the government's two-word form when describing policy.
Is this site's SI Score a measure of how close we are to superintelligence?
No. The SI Score ranks today's released models against each other on published benchmarks. It says nothing about when or whether the research-sense superintelligence will arrive — no leaderboard can measure that.