By the time GPT 5.5 landed we were already on 5.1, honestly they seem to converge on similar limitations around compositional reasoning.