According to Dynamic AI Beating monitoring, Romain Huet, the Developer Experience Lead at OpenAI, revealed during an AMA session that starting from GPT-5.4, the Codex model line (previously a separate series of programming-specific models maintained independently by OpenAI, distinct from the desktop app of the same name) has been merged with the core model into a single system and will no longer iterate separately. Building on this integration, GPT-5.5 has seen further improvements, particularly in agentic coding (agent-driven autonomous programming) and computer use (manipulating computers to complete tasks).
The practical impact of this merger on developers is that they no longer need to choose between a general-purpose model and a programming-specific model, as a single model now covers conversational, reasoning, and programming abilities.
