Purpose: Corporate-startup collaboration is an opportunity for inter-organizational learning. This paper aims to develop an empirically grounded typology to guide researchers and managers in choosing a model that is coherent with the underlying learning processes. Design/methodology/approach: The empirical research consisted of three phases. First, 30 large companies were interviewed to identify different models of interaction. Second, eight different models and cases were selected, and a list of key characteristics of each model was drawn based on the empirical data, obtaining a first typology. Third, the typology was tested, improved and validated on nine corporations. Findings: The main result of the study is an empirically grounded typology of models for corporate-startup interaction and inter-organizational learning. Six dimensions distinguish each model from the others. Research limitations/implications: The paper contributes to extend the literature on inter-organizational learning through different models of corporate-startups interaction. It also contributes to organizational ambidexterity theory, showing how collaboration with startups can improve exploitation and exploration. Practical implications: The typology provides an instrument for analyzing and selecting operational models for corporate-startup collaboration and learning. The models can be modified or broadened, creating new hybrids. Originality/value: The paper presents a novel way of looking at corporate-startup relations by studying them through the lens of organizational learning theory. It explains the existence of different inter-organizational arrangements and provides guidance in selecting the correct model.