[Dataset] Lunara Art Eval: 8,000 Generated Images Across 8 Models, With Prompts And Evaluation Scores

Following the first two releases that reached frontpage of Hugging Face next to the big names, sharing Lunara Art Eval, a public dataset from Moonworks for evaluating artistic quality and prompt adherence in text-to-image generation.

• 8,000 generated images by Moonworks Lunara and other models
• Evaluation scores for aesthetic quality, emotional resonance, and content integrity.

The models evaluated are Lunara, GPT-Image-1 Mini, Qwen-Image, AuraFlow, SD 3.5 Turbo, HiDream-I1 Fast, FLUX-Klein-4B, and Z-Image-Turbo.

This could be useful for comparing model outputs on shared prompts and exploring differences across artistic styles. The dataset could also be used to train loras in different styles. Would appreciaate feedback!

Dataset: https://huggingface.co/datasets/moonworks/lunara-art-eval

Paper: https://arxiv.org/abs/2609.22272