The answers to some of the biggest questions in science may already be hiding inside this dataset
Humanity has long studied the cosmos from within it, unable to step outside and run controlled experiments on the universe itself. This week, the FLAMINGO project—led by Leiden University and powered by the COSMA8 supercomputer—released 2.5 petabytes of simulated cosmic history to the global research community, offering something rare: a virtual universe spanning billions of light-years, freely available to any scientist willing to ask a question of it. In making this digital twin of existence openly accessible, the collaboration has transformed a problem of isolation into one of shared inquiry, placing the deepest unsolved questions in physics—dark matter, dark energy, the geometry of spacetime—within reach of minds that might never have had the tools to approach them.
- Cosmologists have always faced a fundamental constraint: the universe cannot be paused, rewound, or experimentally altered—leaving simulation as the only laboratory available.
- The FLAMINGO dataset, at 2.5 petabytes, models not just galaxies but the vast cosmic web connecting them, capturing structures so immense that no single telescope survey could ever fully map them.
- Building this simulation required the COSMA8 supercomputer and custom-written code called SWIFT—infrastructure so demanding it would have been inaccessible to most research teams working independently.
- Rather than confining results to a closed collaboration, the team built an open platform allowing researchers worldwide to query and download only the data they need, sidestepping the logistical burden of files that would overwhelm most institutions.
- Already informing dozens of studies since its 2023 introduction, the full public release is expected to sharply accelerate discoveries about dark matter, dark energy, and the fundamental shape of the cosmos.
Cosmologists face a problem unique among scientists: they study a system they are trapped inside, with no way to run a controlled experiment on the universe itself. Their only recourse is to build a replica—a digital twin of cosmic history—and test whether the simulated universe their equations produce matches the one their telescopes actually observe.
This week, the FLAMINGO project, led by Leiden University in the Netherlands, made that replica freely available to researchers everywhere. The release comprises 2.5 petabytes of simulated cosmos—equivalent to roughly half a million high-definition films—open to anyone with a research question and an internet connection.
What FLAMINGO models is staggering in scope: billions of light-years of simulated space, capturing not just individual galaxies but the cosmic web—the filamentary network of gas and dark matter along which galaxies are strung like beads. How this structure grew from quantum fluctuations moments after the Big Bang, and how it relates to the dark matter and dark energy comprising 95 percent of all existence, remains among the deepest unsolved problems in physics.
Producing the simulation demanded the COSMA8 supercomputer in the United Kingdom—equivalent in power to seventeen thousand home computers—along with custom code called SWIFT to manage the complexity of modeling billions of galaxies and the invisible matter surrounding them. Rather than keeping these results within a closed collaboration, the team built a dedicated platform allowing scientists to explore and download only the portions they need, sparing institutions the burden of managing impossibly large files.
Since the simulations were first introduced in 2023, they have already shaped dozens of studies on galaxy formation and matter distribution. The full public release is expected to accelerate that pace considerably—and somewhere inside those 2.5 petabytes, the answers to questions that have eluded physics for decades may already be waiting.
Cosmologists face a peculiar problem that no other scientist must confront: they study a system from inside it, with no way to step outside and observe the whole. You cannot run an experiment on the universe itself. You cannot rewind it, adjust a variable, and watch what happens. The only recourse is to build a replica—a digital twin of cosmic history—and see whether the simulated universe that emerges from your equations matches the one your telescopes actually see.
This week, an international team of astrophysicists made the results of their effort freely available to researchers everywhere. The FLAMINGO project, led by scientists at Leiden University in the Netherlands, has released 2.5 petabytes of simulated cosmos—roughly the equivalent of half a million high-definition films, all of it open to anyone with a research question and an internet connection.
The scale of what FLAMINGO models is almost incomprehensible. The simulations span billions of light-years, large enough to capture not just individual galaxies but the vast structures that dwarf them: galaxy clusters containing hundreds or thousands of galaxies each, and the filamentary architecture that connects everything together. Cosmologists call this the cosmic web—a network of gas and dark matter threading through space, with galaxies strung along it like beads on a string. Understanding how this structure grew from the quantum whispers present moments after the Big Bang, and how it relates to the dark matter and dark energy that comprise 95 percent of all existence, ranks among the deepest questions in physics.
Building such a simulation requires computational power beyond what most research groups can access independently. FLAMINGO ran on COSMA8, a supercomputer in the United Kingdom with the processing capacity of seventeen thousand home computers. The team wrote specialized code called SWIFT to manage the staggering complexity of modeling billions of galaxies and the invisible matter surrounding them. This infrastructure alone would have been out of reach for most researchers working alone.
But that is precisely why the team chose to release everything. Rather than hoarding the results within a closed collaboration, they built a dedicated online platform where scientists can explore and download only the portions of data they need, avoiding the logistical nightmare of managing files so large they would overwhelm most institutions. Since the simulations were first introduced in 2023, they have already informed dozens of studies on how galaxies form and how matter distributes itself across cosmic time. The full public release is expected to dramatically accelerate that pace.
For astronomers wrestling with dark matter, dark energy, or the fundamental geometry of spacetime itself, access to a virtual universe of this resolution and scope is not a luxury—it is transformational. The data sits there now, waiting. Somewhere inside those 2.5 petabytes may be the answer to questions that have eluded physics for decades, waiting only for the right researcher with the right idea to find it.
Notable Quotes
The Universe is not a laboratory you can peer into from above, it's the thing you are already inside.— Source material on the fundamental challenge of cosmology
The Hearth Conversation Another angle on the story
Why does a cosmologist need a simulation at all? Can't you just look at the universe and measure things?
You can measure what you see, but you can't test your ideas about how it all works. You can't rewind the universe or run it twice with different rules. A simulation lets you do that—you start from the Big Bang, let the physics play out, and see if what you get matches reality.
So FLAMINGO is just one simulation, then? Why is it such a big deal?
It's not just the size, though 2.5 petabytes is enormous. It's that it models the universe at scales where you can see both individual galaxies and the cosmic web that connects them all. Most simulations have to choose one or the other. FLAMINGO does both.
And now anyone can use it?
Yes. They built a platform so researchers anywhere can download just the pieces they need. Before this, you'd need access to a supercomputer and years of computing time. Now a graduate student with a good question can explore a virtual universe.
What kinds of questions could they answer?
Anything about how dark matter shaped galaxies, how dark energy affects cosmic structure, why the universe looks the way it does. The data is there. The discoveries are waiting for someone to ask the right question.