MolCRAFT is a novel generative model designed for structure-based drug design (SBDD) that operates in the continuous parameter space. The model addresses key challenges in current SBDD models, such as mode collapse in autoregressive methods and the hybrid continuous-discrete space in diffusion models. MolCRAFT introduces a unified SE-(3) equivariant generative model and a novel noise-reduced sampling strategy, which allows for more stable 3D structures and higher binding affinity. Empirical results show that MolCRAFT consistently achieves superior performance in binding affinity with more stable 3D structures, outperforming other strong baselines by a wide margin. To the best of our knowledge, MolCRAFT is the first model to achieve reference-level Vina Scores (-6.59 kcal/mol) with comparable molecular size, demonstrating its ability to accurately model interatomic interactions. The code for MolCRAFT is available at <https://github.com/AlgoMole/MolCRAFT>.MolCRAFT is a novel generative model designed for structure-based drug design (SBDD) that operates in the continuous parameter space. The model addresses key challenges in current SBDD models, such as mode collapse in autoregressive methods and the hybrid continuous-discrete space in diffusion models. MolCRAFT introduces a unified SE-(3) equivariant generative model and a novel noise-reduced sampling strategy, which allows for more stable 3D structures and higher binding affinity. Empirical results show that MolCRAFT consistently achieves superior performance in binding affinity with more stable 3D structures, outperforming other strong baselines by a wide margin. To the best of our knowledge, MolCRAFT is the first model to achieve reference-level Vina Scores (-6.59 kcal/mol) with comparable molecular size, demonstrating its ability to accurately model interatomic interactions. The code for MolCRAFT is available at <https://github.com/AlgoMole/MolCRAFT>.