TDCOSMO measures H0 at 71.1 km/s/Mpc with 3.3% uncertainty using strong lensing and JWST kinematics

A new analysis from the TDCOSMO collaboration puts the Hubble constant at 71.1 kilometers per second per megaparsec, with a 3.3% uncertainty, making it one of the most precise independent late-Universe checks yet on one of cosmology’s biggest disputes.

The result, posted Oct. 10 in the preprint “TDCOSMO 2026: A 3% measurement of the Hubble constant from strong lensing and stellar kinematics” and submitted to Physical Review D, comes from a method known as time-delay cosmography. It uses the delayed arrival times of light from multiple images of distant quasars that have been split and magnified by foreground galaxies.

Why it matters: the Hubble constant describes how fast the universe is expanding today, and different methods have not converged on a single answer. TDCOSMO’s new value lands between two of the best-known benchmarks — the Planck cosmic microwave background estimate of 67.4 ± 0.5 and the H0DN local-consensus value of 73.50 ± 0.81 kilometers per second per megaparsec. That makes the new result an important independent test, but not a resolution of the broader tension.

The study, by TDCOSMO and co-authors including corresponding authors Aymeric Galan, Martin Millon and Pritom Mozumdar, is based on 11 strongly lensed quasars with measured time delays. In its baseline analysis, the team combined those lenses with a prior on matter density from the DES-Dovekie supernova sample and, in flat Lambda-CDM cosmology, found H0 = 71.1 (+2.4/-2.3).

As the abstract puts it: “In the flat Λ cold dark matter (ΛCDM) cosmology, we find H0=71.1^{+2.4}_{-2.3} km.s^-1.Mpc^-1, a 3.3% uncertainty that accounts for the mass-sheet degeneracy.”

That last point is central. The mass-sheet degeneracy is a long-standing ambiguity in strong-lensing models that can bias Hubble-constant measurements if not handled properly. The paper presents the new result as an improved blind analysis relative to the collaboration’s 2025 work, with several upgrades aimed at tightening control of those systematics.

Among them: nine of the 11 lens systems now have spatially resolved stellar kinematics, meaning the team could use detailed measurements of how stars move within the lensing galaxies rather than relying as heavily on population-level assumptions. Six of those resolved kinematic data sets came from JWST’s NIRSpec instrument, and for three lens models, JWST NIRCam imaging replaced earlier Hubble Space Telescope imaging.

The team also replaced an earlier correction-factor approach built on spherical models with axisymmetric dynamical modeling fitted directly to the resolved kinematic data. And for estimating the effect of matter along the line of sight, the study used Euclid Flagship simulations instead of the older Millennium Simulation.

The paper says the shift toward direct kinematic information is significant: “Nine out of 11 systems now have spatially resolved kinematics. Thus, while the external samples remain important for constraining the population-level properties of the deflectors, our hierarchical inference is increasingly informed by detailed kinematic information from the time-delay lenses themselves.”

TDCOSMO also reports consistency checks. Replacing the DES-Dovekie prior with the DESI-DR2 baryon acoustic oscillation likelihood gave “fully consistent” Hubble-constant results, according to the paper. And when the team added the external galaxy samples SLACS and SL2S, which it says are in “excellent agreement” with the time-delay lenses, the precision improved to 3.0%, with H0 = 71.3 (+2.2/-2.1).

For now, though, the headline is not that the Hubble tension is over. It is that a major independent late-Universe method has pushed to roughly 3% precision while explicitly including one of its hardest systematics.

The study is a preprint, arXiv:2610.14781v1, not a peer-reviewed paper yet. The authors say the data and likelihoods will be made public upon acceptance.