The algorithmic intelligence quotient test (AIQ test) is a reasonably well-founded general test of intelligence that is also practically feasible. Deep Q-learning (DQL) and dual-network deep Q-learning (DualDQL) are model-free off-policy deep reinforcement learning agents capable of dealing with complex environments. An experiment with the AIQ test is conducted that evaluates DQL and DualDQL and compares them to the tabular Q-learning. While the agents reach similar AIQ given sufficient training times, for short training times the deep agents outperform the tabular implementation. A hyperparameter search suggests that DQL is more sensitive to its parameters than DualDQL. As their results and resource consumption are otherwise tied, this confirms that DualDQL is the more powerful agent. An initial analysis of the results by environment program length confirms that short programs contribute greatly to the AIQ score, yet the score is not dominated by them.