Message from April 23, yesterday, the developer David AGMM wrote in Reddit that it was open Codex , there is an unexpectedly large unpublished list of models in the drop-down menu。
According to exposure maps and videos, the leak involved the following models:
GPT-5.5 and oai-2.1 are labelled as "the front-line AI smart body"Programming ModelI'm sorry
Arcanine is described as "the front-line model of legendary appetite for starch food."
glocier-alpha is labeled as "wielding the continent's wisdom", while glacier-alpha-block-cy3 is "cold wisdom"
In addition, Heisenberg, described as the "Model of Recent Frontiers of Life Science Research"。
However, if the request is submitted after the GPT-5.5 is selected, Codex returns 400 errors and hints are not supported。OpenAI The model entry was quickly repaired when the problem was discovered and the relevant model entry was downlined from the Cordex interface。