Dopamine is a research-oriented RL framework that provides compact, battle-tested implementations of popular reinforcement learning algorithms including DQN, C51, Rainbow, IQN, SAC, and PPO. It was designed to enable easy experimentation and flexible development for RL research, with implementations available in both JAX and legacy TensorFlow. The framework supports Atari and Mujoco environments and prioritizes reproducibility following established RL best practices.