Didn't expect the pricey GPT-5.2 Pro to have such high usage—its transaction volume has already reached half that of the GPT-5.2 Standard version, but the core user groups for the two are completely different. The fields of science, finance, and law clearly lean more towards paying for the Pro version, while programmers and academia still primarily use the Standard version. Maybe it can be understood this way: coding errors can be iterated on at a low cost, but the tolerance for errors in research and legal fields is much lower, so users are more willing to pay a premium for 'certainty'.