>>105635996
>no API company has ever revealed the size of their API models, so you're asking for the impossible here
I asked for that specifically because the only models that can likely do that perfectly are gpt4o (which by itself is a large multimodal LLM with hundreds of billions of parameters), gemini (same case) and Imagen 3 (made by fucking Google and trained on the entire internet, also the model is likely significantly larger than Flux)