The relationship between vergence and accommodation remains relatively unexplored. The frequent co-occurrence of non-strabismic accommodation and vergence disorders suggests a link between the two processes. Experimental data also reveal that vergence and accommodation have reciprocal effects, where their reflexes influence each other mutually. According to previous findings, chromatic aberration (CA) plays a key role in the learning of accommodation, with the magnitude and direction of defocus being indicated by the relative contrasts detected by the long, medium, and short wavelength cones in the retina. Likewise, vergence control can be acquired from an intrinsic motivation to maximize the encoding efficiency of binocular visual inputs, following the theory of Active Efficient Coding (AEC). We propose a computational model of the development of active binocular vision in which an agent learns to self-calibrate accommodation and vergence control simultaneously. The model is embedded in a reinforcement learning framework in which autoencoders learn an efficient representation of the sensory input that allows the model to implement AEC and contrast values of color channels are derived from the sensory input to exploit CA. Our agent is able to learn near-optimal control policies for vergence and accommodation and achieves correct focus when the state of the environment changes. Furthermore, our model shows correlations between accommodation and vergence in the range of the corresponding experimental data, meaning that the accommodation stimulus influences vergence and vice versa. This model opens up possibilities for future investigations, such as the simultaneous occurrence of accommodation and vergence disorders.