ComfyUI — Models
Quick, concise comparison
Models at a glance
| Model | Native res | Files (size) | Strengths | Best for |
|---|---|---|---|---|
| SD 1.5 | 512×512 | Single .safetensors (~2–4 GB) | Fast, low VRAM, huge LoRA ecosystem | Stylized/anime, drafts, older GPUs |
| SDXL Base 1.0 | 1024×1024 | sd_xl_base_1.0 (~7 GB) | Better detail, composition, text | Photoreal, product shots, higher fidelity |
| SDXL Refiner 1.0 | 1024×1024 (second stage) | sd_xl_refiner_1.0 (~6 GB) | Micro‑details; polish at end of steps | Add after Base when VRAM/time allow |
| Pony (1.5) | 512×512 | Single .safetensors (~2–4 GB) | Anime/cartoon look, strong linework | Stylized/anime, character sheets |
| Pony XL (SDXL) | 1024×1024 | XL checkpoint(s) | Stylized + SDXL fidelity | Higher‑quality anime/stylized renders |
Where to place files
ComfyUI/
models/
checkpoints/ # SD 1.5, SDXL base, SDXL refiner
vae/ # VAEs (if not baked)
loras/ # LoRAs
clip/ unet/ # as needed by some nodes