ABSTRACT Isolated brown dwarfs provide remarkable laboratories for understanding atmospheric physics in the low-irradiation regime, and can be observed more precisely than exoplanets. As such, they provide a glimpse into the future of high-signal-to-noise ratio (SNR) observations of exoplanets. In this work, we investigate several new considerations that are important for atmospheric retrievals of high-quality thermal emission spectra of sub-stellar objects. We pursue this using an adaptation of the h y dra atmospheric retrieval code. We propose a parametric pressure–temperature (P–T) profile for brown dwarfs consisting of multiple atmospheric layers, parametrized by the temperature change across each layer. This model allows the steep temperature gradient of brown dwarf atmospheres to be accurately retrieved while avoiding commonly encountered numerical artefacts. The P–T model is especially flexible in the photosphere, which can reach a few tens of bar for T-dwarfs. We demonstrate an approach to include model uncertainties in the retrieval, focusing on uncertainties introduced by finite spectral and vertical resolution in the atmospheric model used for retrieval (∼8 per cent in the present case). We validate our retrieval framework by applying it to a simulated data set and then apply it to the HST/WFC3 (Hubble Space Telescope’s Wide-Field Camera 3) spectrum of the T-dwarf 2MASS J2339+1352. We retrieve sub-solar abundances of H2O and CH4 in the object at ∼0.1 dex precision. Additionally, we constrain the temperature structure to within ∼100 K in the photosphere. Our results demonstrate the promise of high-SNR spectra to provide high-precision abundance estimates of sub-stellar objects.