Some Notes:

The released benchmarks for GPT 5.5 showed marginal gains; if anything I thought GPT 5.5 might have been more of an improvement on OpenAI's end than the consumer end (providing the same level of outputs with much less thinking tokens and compute power), but after benchmarking them here, I was pretty impressed.

Though again, I can see how people might interpret the results to be quite similar in quality

I will say, with the 5.5 family, the differences between the Pro and standard model are (in my opinion) the least pronounced they've ever been; 5.5 -> 5.5 Pro have very similar output quality

It's uncanny how similar their outputs are actually; I'll likely have to look into adding more difficult/technical prompts; feel free to suggest new ones on the repo

Total cost was $19.98 | Average inference time was: 624 seconds

GPT 5.4 was ~$25 in total; I don't remember the exact cost and unfortunately wasn't documenting costs like I am now

Despite doubling the API costs, OpenAI's claim about the…