This paper investigates model dynamics and risk premia in the short term market for crude oil futures. Stochastic volatility models, with and without jumps, are estimated using data on both futures and option prices. As an economic application we apply the estimated models to the pricing of crude oil variance swaps and an evaluation of the associated variance risk premium. The empirical results point to a positive return risk premium attached to diffusive stochastic volatility while there is not strong evidence of jump risk being priced in the market. Negative volatility and variance risk premia stand out as a robust and significant feature of the data. Jumps play a minor role for representing data and the jump risk component in both variance swaps and variance risk premia is small. Finally, a non-affine model that allows for level dependent volatility of volatility is found to have the best fit to data.