Only think missing from the analysis, and it's a big one, is a comparison on the basis of work done. For a fixed set of real tasks, how much of your quota (or api usage) does it take you to accomplish it.
Because a big part of the price difference is due to Anthropic's models being more expensive that OpenAIs AND using more tokens for the same tasks (at least according to artificialanalysis.ai).
EDIT: That was a delightfully thorough analysis. Nice to see Anthropic taking the value crown, only because Opus 5.5 is such a joy to use.