Dealer is an end-to-end model marketplace that uses differential privacy to enable data exchange between data owners (who receive compensation when their data is used) and model buyers (who buy models trained on data collected by a broker who interfaces with the data owner). Dealer includes a compensation function for data owners based on Shapley values and price sensitivity; and a price function for model buyers based on Shapley coverage sensitivity and noise sensitivity; such that the broker can maximize revenue with arbitrage-free guarantees. The privacy sensitivity of data owners and noise sensitivity of model buyers are measured by the level of differential privacy.