Accurately predicting the flow speed is crucial for applications of coastal ocean circulation simulations such as sediment, larval or contaminant dispersal. This study aims to assess the accuracy of simulated flow speed in a coastal circulation model in comparison with field observations. Deviation between simulated and observed flow speed was assessed in four shallow, coastal locations and four deep, offshore locations in the Gulf of Lion (NW Mediterranean Sea) using six indicators (bias, relative bias, root mean square error, Hanna & Heinold index, correlation and scatter index). Statistical distributions of indicators were calculated during reference periods with low wind, no waves and no stratification. During these periods, relative bias indicated the model displayed a higher performance in predicting transport at shallow stations than at deep stations probably due to grid refinement at these stations. However, there was a low correlation between simulated and observed flow speed, indicating short term time/space mismatches, at all stations during reference periods. Indicators were then calculated during three types of events (wind, waves and stratification) when model assumptions were expected to be violated and their corresponding probability during reference periods indicated that neither wind, wave nor stratification events worsens model’s performance.