A statistician will tell you that all models are wrong, but some are useful.
Particle physicists have an enormously successful model called the Standard Model, which describes the world in terms of seventeen quantum fields, giving rise to particles from the familiar electron to the challenging-to-measure Higgs boson. The model has nineteen parameters, numbers that aren’t predicted by the model itself but must be found by doing experiments and finding the best statistical fit. With those numbers as input, the model is extremely accurate, aside from the occasional weird discrepancy.
Cosmologists have their own very successful standard model that they use to model the universe as a whole. Called ΛCDM, it describes the universe in terms of three things: dark energy, denoted with a capital lambda (Λ), cold dark matter (CDM), and ordinary matter, all interacting with each other via gravity. The model has six parameters, which must be found by observing the universe and finding the best statistical fit. When those numbers are input, the model is extremely accurate, though there have recently been some high-profile discrepancies.
These sound pretty similar. You model the world as a list of things, fix your parameters based on nature, and make predictions. Wikipedia has a nice graphic depicting the quantum fields of the Standard Model, and you could imagine a similar graphic for ΛCDM.
A graphic like that would be misleading, though.
ΛCDM doesn’t just propose a list of fields and let them interact freely. Instead, it tries to model the universe as a whole, which means it carries assumptions about how matter and energy are distributed, and how space-time is shaped. Some of this is controlled by its parameters, and by tweaking them one can model a universe that varies in different ways. But other assumptions are baked in. If the universe had a very different shape, caused by a very different distribution of matter and energy, then we would need a very different model to represent it. We couldn’t use ΛCDM.
The Standard Model isn’t like that. If you collide two protons together, you need a model of how quarks are distributed inside protons. But that model isn’t the Standard Model, it’s a separate model used for that particular type of experiment. The Standard Model is supposed to be the big picture, the stuff that exists and affects every experiment you can do.
That means the Standard Model is supported in a way that ΛCDM isn’t. The Standard Model describes many different experiments, and is supported by almost all of them. When an experiment disagrees, it has specific implications for part of the model only. For example, neutrinos have mass, which was not predicted in the Standard Model, but it proved easy for people to modify the model to fit. We know the Standard Model is not the full picture, but we also know that any deviations from it must be very small. Large deviations would contradict other experiments, or more basic principles like probabilities needing to be smaller than one.
In contrast, ΛCDM is really just supported by one experiment. We have one universe to observe. We can gather a lot of data, measuring it from its early history to the recent past. But we can’t run it over and over again under different conditions, and our many measurements are all measuring different aspects of the same thing. That’s why unlike in the Standard Model, we can’t separate out assumptions about the shape of the universe from assumptions about what it contains. Dark energy and dark matter are on the same footing as distribution of fluctuations and homogeneity and all those shape-related words, part of one model that gets fit together as a whole.
And so while both the Standard Model and ΛCDM are successful, that success means something different. It’s hard to imagine that we find new evidence and discover that electrons don’t exist, or quarks don’t exist. But we may well find out that dark energy doesn’t exist, or that the universe has a radically different shape. The statistical success of ΛCDM is impressive, and it means any alternative has a high bar to clear. But it doesn’t have to mean rethinking everything the way an alternative to the Standard Model would.
4 gravitons 
“When those numbers are input, the model is extremely accurate, though there have recently been some high-profile discrepancies.”
This is far too positive of an assessment. There are dozens of very serious, well replicated discrepancies between ΛCDM and astronomy observations, both at the scale of galaxies and galaxy clusters, and at the scale of cosmology, many of which are independent of each other. Realistically, the only reason that ΛCDM is still being used and remains the paradigm is because there isn’t a consensus around what to replace it with. It is enjoying null hypothesis and seniority privilege.
I summarized many of them, as of 2021, at https://dispatchesfromturtleisland.blogspot.com/2021/01/the-bbc-on-cracks-in-cosmology.html and since then, both the Hubble tension and the Impossible Early Galaxies problem have grown more acute. See, e.g., Sunny Vagnozzi, “Seven hints that early-time new physics alone is not sufficient to solve the Hubble tension” arXiv:2308.16628 (August 31, 2023) (accepted for publication in Universe) and Labbé, I., van Dokkum, P., Nelson, E. et al. “A population of red candidate massive galaxies ~600 Myr after the Big Bang.” Nature (February 22, 2023). https://doi.org/10.1038/s41586-023-05786-2 (Open access version available at https://arxiv.org/abs/2207.12446).
The DESI collaboration has likewise strongly disfavored a constant cosmological constant in recent months. See, e.g., “On DESI’s DR2 exclusion of ΛCDM”arXiv:2504.15336 (LambdaCDM disfavored at 3.1 sigma).A
And, more distinct new problems have emerged since I compiled a list in 2021, like Ziwen Zhang, et al., “Unexpected clustering pattern in dwarf galaxies challenges formation models” arXiv:2504.03305 (April 7, 2025) (Accepted for publication in Nature) and here (the so-called growth index. . . . that the ΛCDM cosmology predicts [is] γ = 0.55. However, using observational data, Ref. Nguyen:2023 measured a much higher γ = 0.633 + 0.025 − 0.024, excluding the ΛCDM value within 3.7σ). Similarly, the dark galaxy “Nube”, the largest low surface brightness galaxy of its kind, shouldn’t happen in the LambdaCDM model either (“Current cosmological simulations within the cold dark matter scenario, including baryonic feedback, do not reproduce the structural properties of Nube”, and also discussing other LamdaCDM problems in the introduction).
And, there is also new stronger evidence like Sergei D. Odintsov, Diego Sáez-Chillón Gómez, German S. Sharov, “Modified gravity/Dynamical Dark Energy vs ΛCDM: is the game over?” arXiv:2412.09409 (December 12, 2024) (“this analysis suggest that standard exponential F(R) models provide much better fits than ΛCDM model, which is excluded at 4σ. Moreover, the parameterisations of the equation of state suggest a non-constant EoS parameter for dark energy, where ΛCDM model is also excluded at 4σ.”).
ΛCDM model is, at best, in the epicycle stage. But inertia and the fact that there are many alternatives on offer and none has swept in to displace it keep it regularly getting mentioned in papers.
Put another way, as a torrent of new astronomy data has rushed in, we are now certain that the simple six parameter ΛCDM model can fit the much more fine grained data set which also contains many specific observations that just can’t work in that model.
The galaxy/galaxy cluster scale solutions must come from the CDM part of the ΛCDM model.
But pretty much the entire parameter space of dark matter particles has been ruled out with results from solid, peer reviewed papers that have been replicated, except ultra-light bosonic dark matter (with particles with a mass on the same order of magnitude as a typical hypothetical graviton, see Oem Trivedi, Abraham Loeb, “On the Cosmological Constant-Graviton Mass correspondence” arXiv:2411.12757 (November 15, 2024)).
Gravitational solutions to the phenomena that CDM describes beyond mere toy-model MOND are less developed and haven’t had their “tires kicked” as much as dark matter particle proposals, but f(R) models, Moffat’s MOG theory, and Alexandre Deur’s approach (whether it is truly just non-perturbative GR as claimed, or is actually a subtle modification of GR), as well as relativistic generalizations of MOND, have made a lot of progress in explaining dark matter phenomena, and sometimes dark energy phenomena as well, although particular versions of these theories have had minor problems. The gravitational models also have the virtue of being far more predictive of new observations that astronomy couldn’t resolve when first hypothesized than ΛCDM.
The death of ΛCDM also probably has as much to do with the sociology of astrophysics and the deaths of its advocates who aren’t up to date on new observational challenges to it, as it does with the merits at this point.
LikeLike
I will say that you’d be a lot more skeptical of this cavalcade of 3 sigma discrepancies in the context of the SM, where it seems like 3 sigma discrepancies happen nine times out of ten. 😉
That said, you’d be right to have that different attitude, which is precisely the point of my post: ΛCDM is intrinsically more vulnerable.
While I do think that there is something sociological going on, I don’t think it’s mostly due to lack of awareness of discrepancies. Rather, when I talk to cosmologists, the issue is that while alternatives tend to do better at handling one discrepancy or another, they have big gaps in handling other parts of the evidence. Sometimes that’s because they contradict that evidence, but other times it’s just because nobody has gotten predictions out of the alternative theory that can be compared to that evidence. The latter is in my view the biggest place where sociology comes in, where ΛCDM is the only proposal that has a community big enough to do something approaching global fits. This should improve over time, but it would be really nice if there were a central database of what’s currently known analogous to the particle data group.
LikeLike
“it would be really nice if there were a central database of what’s currently known analogous to the particle data group.”
Great idea!
It’s easy to complain that individual papers have inadequate reviews of the literature, as they often do. But part of the reason for that is that the literature isn’t well indexed (a problem that is also a perennial issue for searches of prior trademark and patent filings, but with higher stakes).
One of the main reasons I have a science blog of my own is too have a place to gather information that is currently known in the subtopics that interest me.
LikeLike