Context Cereal-legume intercropping offers potential for the sustainable intensification of annual cropping systems. Crop models can support the evaluation of agronomic and environmental benefits of intercropping. However, to be effective, they must capture interspecific interactions during the growing season, a challenging task in tropical systems where such models remain largely untested. Objective This study assessed the ability of the STICS model to simulate cereal-cowpea intercropping productivity and the final outcomes of interspecific interactions occurring over the cropping season under tropical semi-arid conditions. Methods Data from four experimental sites in West Africa, including sorghum, millet, and cowpea grown in sole cropping, were used for parameter calibration. STICS was evaluated using intercropping measurements. To analyse interspecific interactions in intercropping with contrasted sowing patterns, we adapted the partial Land Equivalent Ratio (pLER). Specifically, we corrected the pLER of each crop by dividing it by the ratio of the number of plants per m & sup2; in intercropping to the number of plants per m & sup2; in sole cropping (pLER(cor.dens)). This adjustment allowed us to evaluate interactions at the plant level. Four situations were identified: i) cereal and legume pLER(cor.dens) < 1, indicating that competition outweighed complementarity, compensation, and cooperation or facilitation effects, i.e., productivity of both crops was reduced in intercropping, ii) cereal pLER(cor.dens) > 1 and legume pLER(cor.dens) < 1, indicating that the cereal suppressed legume growth, iii) cereal pLER(cor.dens) < 1 and legume pLER(cor.dens) > 1, indicating that legume suppressed cereal growth and iv) cereal and legume pLER(cor.dens) > 1, indicating that the productivity of both crops was enhanced by intercropping. Results STICS simulated cereal aboveground biomass and grain yield in intercropping with a higher model efficiency (Ef) than 0.48 and cowpea aboveground biomass with Ef = 0.31, but Ef < 0 for cowpea grain yield. The model simulated intercropping yields with low bias and was robust across agro-pedoclimatic conditions, but rRMSE was superior to 30%. In over 70% of irrigation & times; fertilisation & times; site combinations, STICS accurately simulated interspecific interactions, thanks to its accurate simulation of potential yields and water and N limitations. High observed soil moisture at the experimental sites limited the evaluation of model performance in simulating competition for water in intercropping. Conclusion Despite room for improvement, the STICS intercrop model accurately reproduced cereal-cowpea intercropping yields and interspecific interactions in semi-arid tropical conditions.