This paper argues, on the basis of statistical evidence of recent sharp volatility increases in agricultural commodity markets, that such financialized markets should be modeled differently than simple markets for physical goods. The Momagri 2 Model has been designed as a multi-regional modular model. The originality of it lies in the fact that it connects a microeconomic out of equilibrium market module, characterizing agricultural price behavior on the basis of their specific sources of uncertainty, to a macroeconomic computable constrained general equilibrium model. The emphasis is laid on market microstructure, heterogeneity of boundedly rational agents, and the specifics of agricultural markets. Simulations show that under conservative parametric specifications, the integrated model produces strong price volatility of the magnitude and scope experienced in the first decade of the 21st Century. A scenario of markets liberalization increases, rather than decreases, the volatility of commodity prices. We argue that producer's boundedly rational expectations as well as the ones of short term investors, together with market's structures and risk attitude account for the largest part of this volatility. It is suggested that the modular approach used in this paper be applied to all “financialized” markets. Some characteristics of it should indeed also apply to general economic modeling.