We develop an equilibrium concept coined Endogeneous Quantal Response Equilibrium (EQRE) based on heterogeneous players and endogeneous learning in a logistic quantal choice model. Each player has an asymptotically con- sistent estimate of his rival’s rationality index and is able to choose his own rationality level according to a cost-benefit tradeoff. This approach allows to enrich bounded rationality models by incorporating heterogeneous skills and by bridging the gap between stylized facts on the rationality index dynamics and a learning dimension.