Generative models hold great promise for molecular discovery, significantly increasing the size of search space compared to traditional screening libraries. However, most existing methods for small molecule generation suffer from low synthesizability of produced compound candidates, making translation to experimental validation difficult. In this paper, we propose an extension of the GFlowNet framework that operates directly in the space of chemical reactions, offering out-of-the-box synthesizability, while maintaining comparable quality of generated candidates. We demonstrate that with the proposed set of reactions and fragments, it is possible to obtain a search space of molecules orders of magnitude larger than existing screening libraries while offering low costs of synthesis. We also show that the approach scales to very large fragment libraries, further increasing the number of potential molecules. Our experiments showcase the effectiveness of the proposed approach across a range of oracle models, including pretrained proxy models and GPU-accelerated docking.