Per the code snippet here,
capability, must be one of ['TEXT_TO_TEXT', 'IMAGE_TEXT_TO_TEXT', 'EMBEDDING', 'RERANK'] for the sdk's list_models request.

However the result of list models does not adhere to the same constraints in the cli (v3.71.0):
oci generative-ai model-collection list-models
returns various models with capabilities such as "CHAT" or "TEXT_EMBEDDING" which are from the previous version of the sdk (2.150.3 -> 2.163.1).