That's because OpenAI, Anthropic and Google work on many models in parallel which work cooperatively from the user's POV. So GPT-5.6 is just a checkpoint of their multi-model development.