Rapidata reposted this
No one actually knows if a model is good. OpenAI just released their new 4o image model (wtf is that naming, just call it Dall-E 4 already) The only way to truly judge a genAI model is with shit tons of humans. Luckily I am in the "shit ton of humans" business. In just one day we collected data from 200'000 humans on how this new models stacks up. We present the first ever independent large scale benchmark of the new model! They crushed! Black Forest Labs Flux model has finally been dethroned!
Why is deepseek at the lowest? Don't get me wrong, I like OpenAI too
The Imagen3 model is more successful than both Flux and Dalle-3.
I'm testing the model's quality, and it's truly impressive. Thanks for sharing the data Jason Corkill
> wtf is that naming, just call it Dall-E 4 already Different architecture. 4o is multimodal and not a successor of Dall-E 3
Thanks for sharing, impressive how their still leading the field
I'm going to suggest you rebrand your category. (joking aside - awesome work)
Still just pretty pictures. I don't count this as 'intelligence'. Although im well aware this Statement might not age well... 😇
The collected data can be examined further on huggingface https://huggingface.co/datasets/Rapidata/OpenAI-4o_t2i_human_preference
Founder & CEO @ Rapidata | Instant human intelligence for AI at scale
3wCheck out the full results here: https://www.rapidata.ai/leaderboard/image-models George Cameron Barkley Dai Darius Lam Frederic Boesel Artificial Analysis