text-moderation-latest

text-moderation-latest was the alias for OpenAI's text-only Moderation endpoint model. It has been shut down. OpenAI's deprecations page lists text-moderation-latest, text-moderation-stable and text-moderation-007 with a shutdown date of October 27, 2025 (announced April 28, 2025) and omni-moderation as the recommended replacement. The current Moderation guide (the catalog website link, platform.openai.com/docs/guides/moderation, redirects there) documents only omni-moderation-latest.

Key Features

• Replacement: omni-moderation-latest accepts both text and images. Audio is not supported, according to the guide.
• Free to use: the guide states "The moderation endpoint is free to use."
• Categories in the current model: text-only categories are harassment, harassment/threatening, hate, hate/threatening, illicit, illicit/violent and sexual/minors. Text-and-image categories are self-harm (with intent and instructions), sexual, violence and violence/graphic.
• Limits: image files can be up to 20 MB, and the guide warns that category scores may change as the underlying model is upgraded, so custom policies may need recalibration.
• Not for CSAM handling: the guide says not to send known or suspected child sexual abuse material to the Moderation API.

Use Cases

• Screening user prompts and model outputs for policy categories.
• Understanding older safety pipelines that called text-moderation-latest.
• Planning a migration to the omni model and re-tuning thresholds.

Limitation: requests to the retired text models no longer work, and score distributions differ between model generations, so old thresholds should not be copied across without testing.

Pricing and availability

The text-moderation models are shut down. The Moderation endpoint is documented as free to use with omni-moderation-latest; see omni-moderation-latest.

Getting Started

1. Search your code for text-moderation-latest and text-moderation-stable.
2. Switch the model to omni-moderation-latest.
3. Re-evaluate your thresholds on a labelled sample from your own traffic.

FAQ

Is text-moderation-latest still available?

No. OpenAI listed a shutdown date of October 27, 2025.

What replaced it?

OpenAI's recommended replacement is omni-moderation; the current alias is omni-moderation-latest.

Does the replacement support images?

Yes, for the text-and-image categories. Audio is not supported.

Is the Moderation API free?

The guide says the moderation endpoint is free to use.

Alternatives

• omni-moderation-latest: OpenAI's current moderation model.
• Llama Guard 3: Meta's open-weights safety classifier.
• OpenAI: the company page.

Conclusion

text-moderation-latest is a retired alias. Move to omni-moderation-latest and recalibrate your thresholds. See more in the models category.