Résumé : In this talk, I will present MFG models in which the players only have a partial information on the state of the game, namely on the repartition of the other players. This question, which is natural in lots of applications, has been almost unstudied for the moment. I will present tools to tackle such problems. First I will consider a case in which the players do not gain any information during the game. Then,  I will consider cases in which the players will gain information by observing their costs.