A thorough analysis of the uses of “intelligence” approaches to model the nanofluid thermal conductivity is carried out in this study. According to this review, the “predictive models” accuracy is correlated with their structure, used functions, the selected input variables, and the utilized algorithm. For example, compared to statistical correlations achieved by “curve fitting,” ANNs are more detailed. It is also concluded that their output can be noticeably affected by the neural network structure, including the hidden layer’s number and neuron number. The “trial-and-error” process is used to distinguish the most favorable configuration of ANNs. Owing to the reliance of the models on the input component, the greater “thermal conductivity” of the models results in consideration of all the variables influencing the “precision” of the nanofluid. To provide a detailed resource for future research, the considered input variables, the consistency of the proposed models, the type of nanofluids, and the type of implemented model are evaluated and listed in this report.