OpenAI Moderation API
OpenAIOpenAI's Moderation API is a free hosted endpoint that classifies text and images against OpenAI's content-policy categories, including hate, harassment, self-harm, sexual content and violence. It does not classify audio. Moderation scores can also be requested inline with a Responses or Chat Completions generation instead of through a separate call.
Openness
1 high confidence- service
- proprietary hosted API(no weights, no self-host)
- access
- free moderation endpoint
- source
- closed
- license
- proprietary
A closed classifier offered only as a hosted OpenAI API endpoint. No model weights, source code or self-hostable build is released, and use falls under OpenAI's API terms.
- https://developers.openai.com/api/docs/guides/moderation.md recorded 2026-09-29
The markdown edition of OpenAI's moderation guide describes a hosted endpoint: "Use OpenAI moderation models to detect harmful content in text and images" through "the moderation endpoint" or alongside a generated response, and "The moderation endpoint is free to use". Searched the body for open source, self-host, download, weights, license and Hugging Face: none.
- https://developers.openai.com/api/reference/resources/moderations.md recorded 2026-09-29
The API reference documents the endpoint as a single hosted call, post /moderations, which "Classifies if text and/or image inputs are potentially harmful." and takes a model parameter listing omni-moderation-latest and the older text-moderation models.
- https://platform.openai.com/docs/guides/moderation recorded 2026-09-29
The old platform.openai.com guide URL now redirects to developers.openai.com and renders the same guide, which says "Use OpenAI moderation models to detect harmful content in text and images" and "The moderation endpoint is free to use".
Adoption
4 medium confidenceOpenAI publishes no request volume, developer count or customer count for this endpoint. The reading instead reflects standing: this is the free default moderation filter available to OpenAI's entire developer base, a credible claim but not a measured one.
- https://developers.openai.com/api/docs/guides/moderation.md recorded 2026-08-13
Guide documents the endpoint and states it is free to use. No usage figure of any kind appears on the page.
Capability
3 low confidenceThe endpoint scores text and images against thirteen fixed categories and accepts no custom policy, unlike policy-conditioned models that reason over a policy the user writes. It does not classify audio.
- https://developers.openai.com/api/docs/guides/moderation.md recorded 2026-09-29
The guide says the omni-moderation-latest "model accepts text and image inputs. It doesn't classify audio." Its category table lists thirteen fixed categories, from harassment and hate through illicit, self-harm, sexual and violence, with "Categories marked as" text only not supporting images, and warns that "custom policies that rely on" category scores "may need recalibration over time"; there is no option for supplying a custom policy.
- https://developers.openai.com/api/reference/resources/moderations.md recorded 2026-09-29
The API reference documents the endpoint as a single hosted call, post /moderations, which "Classifies if text and/or image inputs are potentially harmful." and takes a model parameter listing omni-moderation-latest and the older text-moderation models.
Verified 2026-09-29, except some scores, still awaiting re-confirmation