An agent facing uncertain losses over a finite period aims to optimize a mean-variance criterion with multiple loss distribution models through risk sharing. Adapting Maccheroni et al.'s (2009) monotone mean-variance preferences, the study constructs a criterion for multiple models, using dual representation to ensure time-consistency. Assuming a Cramér-Lundberg loss model, the optimal risk-sharing contract and wealth process are characterized, with the strategy proven admissible and the value function verified. Applied to a Spanish automobile insurance portfolio, the strategy uses cross-validation to derive models, with numerical results illustrating the outcomes.