OpenAI updates Model Spec; intended behavior is not guaranteed behavior
The revised specification clarifies human intent and oversight, without a synchronized product rollout.
On October 29, OpenAI revised its Model Spec, clarifying how models should interpret context and preserve human oversight while retiring most detailed style rules.
This describes intended behavior. OpenAI says current models do not fully reflect it; it does not guarantee simultaneous changes across ChatGPT or the API. Not independently tested.
https://model-spec.openai.com/2026-10-29.html
This describes intended behavior. OpenAI says current models do not fully reflect it; it does not guarantee simultaneous changes across ChatGPT or the API. Not independently tested.
https://model-spec.openai.com/2026-10-29.html
Usage conditions
- Hong Kong
- Public specification, not a local product launch. The English page showed no reading paywall. This announcement does not establish new web/app/API features, plans, Traditional Chinese versions or rollout timing in these markets.
- Macau
- Public specification, not a local product launch. The English page showed no reading paywall. This announcement does not establish new web/app/API features, plans, Traditional Chinese versions or rollout timing in these markets.
- Taiwan
- Public specification, not a local product launch. The English page showed no reading paywall. This announcement does not establish new web/app/API features, plans, Traditional Chinese versions or rollout timing in these markets.
Official source: An updated Model Spec for more capable and beneficial AI