OpenAI rewrites its Model Spec: ChatGPT and API behavior now rests on five values

On 29 October 2026 OpenAI published an updated Model Spec, the document describing how its models should behave. It is now organized around five values, most style rules are retired, and the assistant is called GPT in the text.

On 29 October 2026 OpenAI published an updated version of the Model Spec, the document that publicly describes how ChatGPT and API models should act. According to the official post the text is now organized around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.

OpenAI says the values summarize principles the document already contained but had spread across many detailed rules. Most detailed style rules are retired, and more behavior is treated as a default that a developer or user can adjust. The character the rules describe is no longer called the assistant in the text but GPT.

OpenAI also acknowledges that its models do not yet fully reflect the new Model Spec. So it describes a target, not a guarantee of current behavior. Independent secondary sources agree on the five values and the renaming, but they conflict on details about agent autonomy, so we leave those out.

For companies building their own products on OpenAI models the document is useful for one reason: it shows what you can reasonably expect from the model and what you cannot. System instructions and behavior rules in your application work as settings on top of these defaults, so they need to be written deliberately and tested.