I'm only half convinced that DALL-E 3.5 is using actual GPT-4 and not a proprietary model

I've been using this thing for almost a month now, and I've had experiences that just don't add up to what I know about GPT-4. It often struggles with context, failing to pick up subtle details in the prompt, and has seemingly inconsistent knowledge about real-world facts.<br><br>If you use the new Playground, you should have noticed by now that GPT-4.5 is almost on a completely different level of intelligence. I can't see DALL-E 3.5 using the later in terms of generative capabilities.<br><br>Since DALL-E 3 was released, I have a feeling that we are being sold manure in diamond packaging.