Temperature-dependent differences between empirical lithium age estimators in benchmark open clusters

Abstract The two public versions of the EAGLES empirical lithium age estimator, analytic (v1) and neural-network (v2.0), are applied to identical published lithium equivalent widths for Blanco 1, the Pleiades and NGC 2516. In these samples, they agree within their combined 68 per cent uncertainties above 4900 K and within half those uncertainties in the three tested windows above 4700 K. For the cool subsets (at or below 4300 K) the ages differ by 37 to 136 Myr. In Blanco 1, where 2 of 103 cool members have lithium detections above three sigma, the versions agree on detected stars at every tested temperature boundary; this selection does not establish the cause of the difference. In the Pleiades cool members and the NGC 2516 K dwarfs they differ on detected stars too. Extending the cool sample to warmer stars yields agreement only in the Pleiades. The external ages do not settle which version is right: two clusters belong to both calibrations, the Blanco 1 depletion-boundary ages favour the network, and an asteroseismic age for NGC 2516 favours the analytic model for the original samples. For cool-dominated samples in this 60 to 300 Myr comparison, users should report the estimator spread.