Has AlphaFold 3 achieved success in predicting RNA 3D structures? This review evaluates AlphaFold 3's performance in RNA structure prediction. RNA structure prediction remains a challenge due to fundamental differences between RNA and proteins. AlphaFold 3, which originally focused on proteins, has been extended to predict structures of DNA, ligands, and RNA. It uses sequence and multiple sequence alignment as inputs and employs a multi-cross diffusion model to predict atom coordinates directly. While AlphaFold 3 shows promise, it does not yet match the success of protein prediction. The article compares AlphaFold 3 with ten existing methods, including ab initio, template-based, and deep learning approaches. Results show that AlphaFold 3 performs well in some metrics but has limitations in predicting non-Watson-Crick interactions and long RNAs. The results are available on the EvryRNA platform. The study highlights the challenges in RNA structure prediction and the current state of AlphaFold 3 in this area.Has AlphaFold 3 achieved success in predicting RNA 3D structures? This review evaluates AlphaFold 3's performance in RNA structure prediction. RNA structure prediction remains a challenge due to fundamental differences between RNA and proteins. AlphaFold 3, which originally focused on proteins, has been extended to predict structures of DNA, ligands, and RNA. It uses sequence and multiple sequence alignment as inputs and employs a multi-cross diffusion model to predict atom coordinates directly. While AlphaFold 3 shows promise, it does not yet match the success of protein prediction. The article compares AlphaFold 3 with ten existing methods, including ab initio, template-based, and deep learning approaches. Results show that AlphaFold 3 performs well in some metrics but has limitations in predicting non-Watson-Crick interactions and long RNAs. The results are available on the EvryRNA platform. The study highlights the challenges in RNA structure prediction and the current state of AlphaFold 3 in this area.