ObjectiveVariants in GABRA1 have been associated with a broad epilepsy spectrum, ranging from genetic generalized epilepsies to developmental and epileptic encephalopathies. However, our understanding of what determines the phenotype severity and best treatment options remains inadequate. We therefore aimed to analyse the electro‐clinical features and the functional effects of GABRA1‐variants to establish genotype–phenotype correlations.MethodsGenetic and electro‐clinical data of 27 individuals (22 unrelated and 2 families) harbouring 20 different GABRA1 variants were collected and accompanied with functional analysis of 19 variants.ResultsIndividuals in this cohort could be assigned into different clinical subgroups based on the functional effect of their variant and its structural position within the GABRA1 subunit. A homogenous phenotype with mild cognitive impairment and infantile‐onset epilepsy (focal seizures, fever sensitivity and EEG posterior epileptiform discharges) was described for variants in the extra‐cellular domain and the small transmembrane loops. These variants displayed loss‐of‐function (LoF) effects and the patients generally had a favourable outcome. A more severe phenotype was associated with variants in the pore‐forming transmembrane helices. These variants displayed either gain‐of‐function (GoF) or LoF effects. GoF‐variants were associated with severe early‐onset neurodevelopmental disorders, including early infantile developmental and epileptic encephalopathy.InterpretationOur data expand the genetic and phenotypic spectrum of GABRA1‐epilepsies and permit to delineate specific sub‐phenotypes for LoF and GoF variants, though the heterogeneity of phenotypes and variants. Generally, variants in the transmembrane helices cause more severe phenotypes, in particular GoF variants. These findings establish the basis for a better understanding of the patho‐mechanism and precision medicine approach in GABRA1‐related disorders. Further studies in larger populations are needed to provide a conclusive genotype–phenotype correlation.This article is protected by copyright. All rights reserved.