DiffClass is a novel exemplar-free class incremental learning (CIL) method that addresses catastrophic forgetting and the domain gap between synthetic and real data. The method uses multi-distribution matching (MDM) diffusion models to align synthetic and real data distributions, reducing domain gaps and improving model stability. It also incorporates selective synthetic image augmentation (SSIA) to enhance training data distribution, boosting model plasticity and the effectiveness of multi-domain adaptation (MDA). By reformulating CIL as a multi-domain adaptation problem, DiffClass implicitly addresses domain gaps during incremental learning, leading to improved performance. Extensive experiments on CIFAR100 and ImageNet100 benchmarks show that DiffClass outperforms existing exemplar-free CIL methods with non-marginal improvements, achieving state-of-the-art results. The method is effective in various CIL settings, demonstrating its robustness in balancing stability and plasticity. The project is available at https://cr8br0ze.github.io/DiffClass.DiffClass is a novel exemplar-free class incremental learning (CIL) method that addresses catastrophic forgetting and the domain gap between synthetic and real data. The method uses multi-distribution matching (MDM) diffusion models to align synthetic and real data distributions, reducing domain gaps and improving model stability. It also incorporates selective synthetic image augmentation (SSIA) to enhance training data distribution, boosting model plasticity and the effectiveness of multi-domain adaptation (MDA). By reformulating CIL as a multi-domain adaptation problem, DiffClass implicitly addresses domain gaps during incremental learning, leading to improved performance. Extensive experiments on CIFAR100 and ImageNet100 benchmarks show that DiffClass outperforms existing exemplar-free CIL methods with non-marginal improvements, achieving state-of-the-art results. The method is effective in various CIL settings, demonstrating its robustness in balancing stability and plasticity. The project is available at https://cr8br0ze.github.io/DiffClass.