The embodiment of the invention provides a dialogue method. The dialogue comprises multiple rounds of dialogues between two parties of the dialogue, one party of the two parties of the dialogue is a target intelligent agent, the other party of the two parties of the dialogue is a user or an opponent intelligent agent, and the method comprises the following steps: estimating an action correspondingto the target intelligent agent according to opponent utterances in the dialogue of this round; estimating an action corresponding to the opponent based on the estimated action; selecting an optimalaction for the opponent action; and generating a corresponding utterance according to the selected optimal action. According to the method, the action of the opponent is estimated to select the more appropriate action and generate the corresponding utterance, so that the decision of the intelligent agent is forward-looking, the more appropriate dialogue can be generated, and better experience is brought to the user. In addition, the embodiment of the invention provides a dialogue device, a medium and computing equipment.