6D object pose estimation is crucial in the field of computer vision. However, it suffers from a significant lack of large-scale and diverse datasets, impeding comprehensive model evaluation and curtailing downstream applications. To address these issues, this paper introduces Omni6DPose, a substantial benchmark featured by its diversity in object categories, large scale, and variety in object materials. Omni6DPose is divided into three main components: ROPE (Real 6D Object Pose Estimation Dataset), which includes 332K images annotated with over 1.5M annotations across 581 instances in 149 categories; SOPE(Simulated 6D Object Pose Estimation Dataset), a simulated training set created by mixed reality and physics-based depth simulation; and PAM(Pose Aligned 3D Models), the manually aligned real scanned objects used in ROPE and SOPE. Omni6DPose is inherently challenging due to the substantial variations and ambiguities. To address this issue, we introduce GenPose++, an enhanced version of the SOTA category-level 6D object pose estimation framework, incorporating two pivotal improvements: Semantic-aware feature extraction and Clustering-based aggregation. Moreover, we provide a comprehensive benchmarking analysis to evaluate the performance of previous methods on this new large-scale dataset in the realms of 6D object pose estimation and pose tracking.