Developmental dyscalculia (DD) is an heterogenous neurodevelopmental learning disability that manifests as persistent difficulties in learning mathematics. DD can occur in isolation but is often diagnosed as a co-occurring difficulty in children with language-based learning disabilities. Basic cognitive and neuroimaging findings suggest different subtypes of dyscalculia exist. However, a comprehensive theoretical framework that provides accepted terminology and clinical criteria to design appropriate interventions is still lacking. We developed a comprehensive battery of cognitive tests, the UCSF Dyscalculia Subtyping Battery (DSB), aiming at identifying deficits in four distinct mathematical domains: number processing, arithmetical procedures, arithmetic facts retrieval, and geometrical abilities. The mathematical abilities of a cohort of 75 children aged 7 to 16, referred to the UCSF Dyslexia Center for a language-based neurodevelopmental disorder, were initially evaluated using a behavioral neurology approach. A team of professional clinicians classified children with difficulties in mathematics in four groups, depending on their parents’ and teachers’ reported symptoms and clinical history, in one of the following domains: number processing, arithmetical procedures, arithmetic facts retrieval and geometrical abilities. The 75 children and 18 typically developing control children were then evaluated with the DSB to identify which subtests of the battery better represented each group. We describe the detailed profiles of four cases, each of them representative of deficits in one of the four domains, and report the pattern of impairment in the overall cohort. Our results show that a neuroscience-based DD evaluation battery enables identification of subtypes acknowledging the multidimensional nature of the disorder. If corroborated in large samples, these findings can pave the way for novel diagnostic approaches, consistent subtype classification, and ultimately personalized interventions.