We’re introducing a new moderation model in the Moderation API today. This model is now multimodal—it supports both text and images. It’s also more accurate (especially for non-English content), can detect harmful content in two new categories (illicit and illicit/violent), and remains free. 
You can read more in our blog post or take a look at our updated documentation: Read the docs 
Our previo...