Neural network model used on research and commercial setting tends to increase in size. This increase in model size led to increase in hardware and energy requirement for design, train, and inference. Increased energy and hardware requirement caused carbon emission that affects environment. This paper performs a review on methods used in three steps of neural network model development: design, training, and inference. The review then discusses on methods’ performance as well as resource usage efficiency. It also summarizes findings and recommends direction for future research effort regarding efficient development of neural network.