AbstractAdvances in analytical methods have made it possible to obtain high-resolution water isotopic data from ice cores. Their spectral signature contains information on the diffusion process that attenuated the isotopic signal during the firn densification process. Here, we provide a tool for estimating firn-diffusion rates that builds on the Community Firn Model. Our model requires two main inputs, temperature and accumulation, and it calculates the diffusion lengths for δ¹⁷O, δ¹⁸O and δD. Prior information on the isotopic signal of the precipitation is not a requirement. In combination with deconvolution techniques, diffusion lengths can be used in order reconstruct the pre-diffusion isotopic signal. Furthermore, the temperature dependence of the isotope diffusion and firn densification makes the diffusion length an interesting candidate as a temperature proxy. We test the model under steady state and transient scenarios and compare four densification models. Comparisons with ice core data provide an evaluation of the four models and indicate that there are differences in their performance. Combining data-based diffusion length estimates with information on past accumulation rates and ice flow thinning, we reconstruct absolute temperatures from three Antarctic ice core sites.