Revised DESI reanalysis suggests unusually tight cosmological bounds on neutrino mass, favors normal ordering

A revised arXiv preprint reports unusually tight cosmological limits on the total mass of neutrinos from a combined reanalysis of public data from the Dark Energy Spectroscopic Instrument, or DESI, and related surveys. The authors say the result amounts to the strongest evidence yet from cosmology for the normal ordering of neutrino masses. But the work is preliminary, has not been peer-reviewed, and the paper itself highlights sensitivities to modeling choices and to tensions among some datasets.

The paper, “Reanalyzing DESI DR1: 4. Percent-Level Cosmological Constraints from Combined Probes and Robust Evidence for the Normal Neutrino Mass Hierarchy,” is an arXiv preprint, identifier arXiv:2601.16165v2, revised July 22, 2026, by Mikhail M. Ivanov, James M. Sullivan, Roger de Belsunce, Shi-Fan Chen, Anton Chudaykin, Mark Maus and Oliver H. E. Philcox. In the standard ΛCDM cosmology model, it reports a 95% confidence upper limit on the sum of neutrino masses of Mν less than 0.049 electron volts. In a broader model with time-varying dark energy, w0waCDM, it finds Mν less than 0.077 eV. The authors say those bounds disfavor the inverted neutrino mass hierarchy at about 3.5 sigma in ΛCDM and about 2.4 sigma in w0waCDM. As the abstract puts it, “The preference for the normal neutrino mass ordering thus holds regardless of the background model.”

Why that matters comes from particle physics. Neutrino oscillation experiments measure differences between neutrino masses, not the full total, but those measurements imply minimum sums of about 0.058 eV for the normal ordering and about 0.098 eV for the inverted ordering. That means a cosmological upper limit below roughly 0.098 eV starts to press directly on the inverted case. By comparison, direct laboratory limits are still much weaker; the paper cites the KATRIN experiment at mβ less than 0.45 eV at 90% confidence.

DESI is a major cosmology survey that maps the three-dimensional distribution of galaxies to track the universe’s expansion and the growth of large-scale structure. In this analysis, the authors use public DESI Data Release 1 galaxy clustering data and combine it with DESI baryon acoustic oscillation measurements, photometric galaxy samples and cross-correlations with cosmic microwave background lensing, which traces matter through the bending of ancient light. Some dataset combinations also include Planck measurements of the cosmic microwave background or the Pantheon+ supernova compilation.

Beyond neutrinos, the paper says the combined analysis sharpens several core cosmological parameters. For its full ΛCDM combination, it reports a Hubble constant, H0, of 69.08 ± 0.37 kilometers per second per megaparsec and a matter density, Ωm, of 0.2974 ± 0.0050. It also gives σ8 = 0.838 ± 0.017 and S8 = 0.834 ± 0.018, two measures related to how matter is clumped across the universe.

The caution flags are important. The paper is not journal-accepted on its arXiv page, and the authors devote space to known issues that can affect neutrino-mass inference, including the Planck AL lensing anomaly, uncertainty in the reionization optical depth and geometric tensions between cosmic microwave background and baryon acoustic oscillation data. They also say some cosmological posteriors for neutrino mass peak at unphysical negative effective values, so they model the marginalized posterior as a Gaussian truncated at Mν = 0. For one baseline combination, they report p(Mν > 0.098 eV) = 0.017, which they translate to about 2.4 sigma. If the result holds up, it would mark a notable step for cosmology’s role in the neutrino mass-ordering question. For now, though, it remains an intriguing and unusually strong preprint claim, not a settled result.