The computational cost of estimating option valuation models is very high. We propose to address these constraints by filtering the state variables using particle weights based on model-implied spot volatilities rather than model prices. We illustrate our method by estimating stochastic-volatility and double-jump models. Using long time series and large cross-sections has important implications for option pricing and asset pricing more generally. The variance risk premium parameter is somewhat smaller and much more precisely estimated. Inference on parameters characterizing skewness, kurtosis, and risk aversion changes. Moneyness and especially maturity restrictions may result in identification problems for the models we study.