Background and objectives: The in vitro gas production test (IVGP) is used for evaluating the nutritional value of feedstuffs in animal nutrition. In IVGP, the volume of gas produced during incubation time is fitted by using a nonlinear model and some parameters related to ruminal fermentation kinetic are predicted. Over the past years, animal nutrition researchers have proposed a variety nonlinear models for this purpose. However, it has been reported that each of these models have predicted ruminal fermentation parameters with different accuracy and presented contradictory results. The purpose of this experiment was to compare the accuracy of some nonlinear models for predicting of gas production kinetics in different diets. Materials and methods: For this experiment, five diets with different forage to concentrate ratio (70: 30, 60: 40, 50: 50, 40: 60 and 30: 70) were prepared. Feedstuffs which used for preparing of these diets included alfalfa hay, wheat straw, barley grain, wheat bran and soybean meal. For investigating the ruminal fermentation kinetic of the experimental diets, the IVGP was used in three separate runs. For each run, rumen fluid was obtained from three fistulated Mehraban rams before the morning feeding. Then, 200 mg of each dried experimental diets with 30 ml of buffered rumen fluid were transferred into each glass vial (in three replications) and all vials were capped. The prepared vials (three vials were considered as the blanks) were incubated at 39 ° C and the volume of gas produced were recorded at 2, 4, 6, 8, 10, 12, 16, 20, 24, 36, 48, 72, 96, 120, 144 hours after incubation. The results (the volume of gas produced at different incubation times) were fitted for exponential (EXP), exponential with lag time (EXPL), Mitscherling (MIT), Korkmaz-Uckardes (KOR) and Weibull (WEB) models. The models goodness of fit were performed using mean square error (MSE), coefficient of determination (R2), residual mean absolute deviation (RMAD), mean percentage error (MPE) and relative efficiency (RE). Durbin-Watson (DW) test and Akaike's information criterions (AIC) were used for selection of the best model. Results: Results showed that the models had no significant difference in term of asymptotic gas volume (A). The EXP model had the highest MSE (35. 74), RMAD (4. 67), MPE (4. 46) and lowest R2 (0. 970), (P < 0. 05). The lowest MSE (2. 63) and RMAD (1. 13) were observed in the WEB model (P < 0. 05). The RE of the WEB model was higher than the other models. The highest (56. 18) and lowest (23. 29) AIC values were observed in the EXP and WEB models, respectively. The DW value in the EXP, EXPL, MIT, KOR and WEB models were 0. 260, 1. 187, 1. 187, 0. 955 and 1. 653, respectively. Conclusion: Generally, results showed that the EXP model predicted gas production kinetic of the tested diets less accurately and the WEB model had the highest accuracy.