Carbon nanotubes (CNTs) are promising materials for neuromorphic devices due to their high mobility, stability, and scalability for various application. Among them, semiconducting CNTs (sc-CNTs) are suitable for analogue conductance modulation, although forming uniform networks remains challenging. Here, we demonstrate synaptic transistors based on high-purity sc-CNT networks. The sc-CNTs enable stable charge trapping and controlled conductance modulation, exhibiting p-type switching with an on/off ratio of ~ 7 × 106 and non-volatile memory behavior. Synaptic functions, including short- and long-term memory as well as long-term potentiation and depression, are emulated with stable operation over 10,000 pulse cycles. Furthermore, low nonlinearity and asymmetry lead to a handwritten image recognition accuracy of 90.26% in artificial neural network simulations. These results indicate that high-purity sc-CNT networks are a reliable platform for high-precision neuromorphic circuitry.