<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/da9ea5aa-91c5-48e4-8afc-15cc33609a26/original=true/da9ea5aa-91c5-48e4-8afc-15cc33609a26.jpeg" />A free, compact, and fast text-to-image model developed by <a rel="ugc" href="https://huggingface.co/AiArtLab/sdxs-2b">AiArtLab</a> for training on consumer-grade GPUs.CosmosTransformer3D: 2b parametersQwen3.5: 0.8b parametersQwen vae: 16ch8xResolution: Default 768x1152, trained from 576 to 1152 with step 64Limitations: trained on small datasets ~2.5kk, focused on art / illustrations / anime and photoCaptions: danbooru, natural (short/medium), trained with 250 max toks<a target="_blank" rel="ugc" href="https://github.com/aiartlab/sdxs-2b-pipeline">Comfyui pipeline</a>

Files with SHA256: c3744a94ff58870ff781fc449d723f73eda935ea755b27891411173bb205013b