Tunnel damages are increasing worldwide in major earthquakes, some of which may result from the underestimated seismic actions and uncertainties associated with the design parameters. Opposite to traditionally performed deterministic analysis, this paper introduces stochastic dynamic time-history analysis to quantify the variability of seismic-induced tunnel deformations in a probabilistic framework. Two-dimensional nonlinear finite difference models combined with the Monte Carlo simulation are employed to assess the influence of parameter uncertainty. Analyses are performed for a wide range of soil shear wave velocities, ground motion intensities, lining Young’s moduli, correlation structures, distribution types, and coefficients of variation. The results indicate that the parameter uncertainty can have a significant impact on tunnel seismic deformations. Increased motion intensity leads to an increased standard deviation, especially for flexible tunnels. The influence of shear strength parameters is insignificant, despite a slightly larger standard deviation appears for the positive correlation. The distribution type effect increases sharply for the cases with high variation degrees, and the lognormal distribution generally predicts a lower mean, standard deviation, and probability of exceedance. The results also show that the soil parameter uncertainty can be properly considered by introducing a factor of safety. These findings may provide a useful reference for the reliability-based seismic design of tunnels.