The Simulation That Couldnt Fix the Cracks in Cosmologys Standard Model

The most detailed simulation of the universe ever made was built to answer a simple question: why does the cosmos appear less lumpy than our best models predict? The answer it got was that the models might be wrong.

FLAMINGO — Full-hydro Large-scale structure simulations with All-sky Mapping — released more than 2.5 petabytes of cosmological data publicly on April 28th, making the full output of one of the largest computational efforts in science available to anyone with an internet connection. The dataset spans 22 hydrodynamical simulations and 16 gravity-only runs, modeling the evolution of matter from shortly after the Big Bang to the present day across volumes of a billion light-years. The comparison to half a million HD movies is accurate. So is the implication that most researchers could never run this on their own hardware.

The simulation was specifically designed to resolve a discrepancy in cosmology called the S8 tension. S8 is a parameter that describes how clumpy the distribution of matter is in the universe. Measurements of the cosmic microwave background — the oldest light in the universe, imprinted 380,000 years after the Big Bang — predict one value for S8. Direct observations of how gravity bends light across billions of years of cosmic history predict a lower one. The universe looks less lumped together than the standard model of cosmology says it should.

FLAMINGO was supposed to close that gap. The hope was that previous simulations, which modeled only dark matter, were missing something: the physics of ordinary matter — gas pressure, supernova explosions, active galactic nuclei blowing material out of galaxies. FLAMINGO includes all of that, tracking dark matter, ordinary matter, and neutrinos simultaneously in a self-consistent framework. It runs on the COSMA8 supercomputer at Durham University using the SWIFT code, distributing work across tens of thousands of CPUs.

What the 2023 FLAMINGO results showed, and what the current data release re-confirms, is that adding ordinary matter does not fix the discrepancy. An exciting possibility is that the tension is pointing to shortcomings in the standard model of cosmology, or even the standard model of physics, Joop Schaye of Leiden University, who leads the project, told Space.com at the time.

That is a cosmologist saying out loud that our best theory of the universe might be broken.

The standard model in question is Lambda-CDM — the Cold Dark Matter model with a cosmological constant for dark energy. It is not a fragile hypothesis. It correctly predicts the large-scale structure of the universe, the cosmic microwave background's detailed structure, the relative abundances of hydrogen and helium from Big Bang nucleosynthesis, and the acceleration of the universe's expansion. If Lambda-CDM has a structural flaw, it does not just affect one niche calculation — it affects every inference astronomers have made about dark energy, galaxy formation, and cosmic history using JWST, Euclid, and the Vera Rubin Observatory.

The counterargument is that FLAMINGO represents one team's simulation with one set of subgrid physics choices. Baryonic feedback processes — how supernovae and black holes redistribute matter — are calibrated against observations we have, and those calibrations carry assumptions. Other teams with different simulation codes might reach different conclusions about whether baryons can account for the S8 discrepancy. The 2023 papers acknowledged this explicitly: the simulations agree very well with observed properties of individual galaxies and clusters, yet still cannot explain the large-scale lumpiness deficit.

Ian McCarthy of Liverpool John Moores University, a co-author, put it this way in 2023: It seems the universe behaved as expected for a significant fraction of cosmic history, but that something changed later on in cosmic history. That is a very careful way of saying something went wrong in the last few billion years. Nobody knows what.

The data release opens the full simulation to external researchers. FLAMINGO's creators have set up a web service at dataweb.cosma.dur.ac.uk allowing selective downloads rather than requiring users to move the entire 2.5-petabyte dataset. Carlos Frenk of Durham University, who worked on the project, noted that most researchers simply do not have access to facilities like COSMA8 — the open release is intended to close that gap. The simulations have already been used in dozens of studies since their 2023 introduction, but the full dataset is new.

The honest version of where the field stands: FLAMINGO validated Lambda-CDM's predictions for galaxy formation in detail. It did not resolve the tension it was built to resolve. The universe is less lumpy than the model predicts, and after running the most sophisticated simulation of cosmic evolution ever attempted, the leading explanation among the people who built it is that the model itself is wrong.

That is either a profound result or the world's most expensive null result. The dataset is now public. The field will decide which.