Abstract
In this paper we examine the capacity of arbitrage-free neural stochastic differential equation market models to produce realistic scenarios for the joint dynamics of multiple European options on a single underlying. We subsequently demonstrate their use as a risk simulation engine for option portfolios. Through backtesting analysis we show that our models are more computationally efficient and accurate for evaluating the value-at-risk of option portfolios than standard filtered historical simulation approaches, with better coverage and less procyclicality.