Table 5. The accuracy of GPT-4 on the GAT Arabic and English questions compared to ChatGPT.

AR and EN refer to Arabic and English respectively. The best performance is indicated in bold.

Question type	EN (GPT-4)	AR (GPT-4)	EN (ChatGPT)	AR (ChatGPT)
Reading comprehension	86.81%	74.29%	80.22%	55.71%
Analogy	73.39%	57.02%	54.03%	37.19%
Contextual error	83.52%	63.37%	68.13%	38.61%
Sentence completion	87.33%	75.47%	82.67%	35.85%
Micro-average	82.68%	67.74%	71.49%	42.74%
Macro-average	82.76%	67.54%	71.26%	41.84%