Gas fuel direct injection (DI) technology can improve the control precision of the in-cylinder mixing and combustion process and effectively avoid volumetric efficiency reduction in a compressed natural gas (CNG) engine, which has been a tendency. However, compared with the port fuel injection (PFI) method, the former’s mixing path and duration are shortened greatly, which often leads to poor mixing uniformity. What is worse, the in-cylinder mixing performance would be seriously affected by engine working conditions, such as engine speed and load. Based on this situation, the fluid mechanics software FLUENT is used in this article, and the computational fluid dynamics (CFD) model of the injection and mixing process in a gas-fueled direct injection engine is established. A quantitative evaluation mechanism of the in-cylinder mixing performance of the CNG engine is proposed to explore the influencing rule of different engine speeds and loads on the mixing process and performance. The results indicate that phase space analysis can accurately reflect the characteristics of the mixture mixing process. The gas fuel mixture rapidly occupies the cylinder volume in the injection stage. During the transition stage, the gas fuel mixture is in a highly transient state. The diffusion stage is characterized by the continuous homogenization of the mixture. The in-cylinder mixing performance is linearly dependent on the engine’s working condition in the phase space.