Qwen-Image-Edit 2509 VRAM Requirements
Qwen-Image-Edit 2509 needs about 20.5 GB of VRAM with FP8 weights at 1024×1024, loading one model at a time. In FP16 it needs 39.5 GB (57.6 GB with everything loaded), and GGUF Q4 brings it to 12.4 GB. The smallest consumer GPU that runs it is the RTX 4060 Ti 16GB with Q4 weights, or the RTX 3090 24GB in FP8.
20.5 GB peak VRAM
Qwen-Image-Edit 2509 at 1024×1024 with FP8 weights, loading one model at a time (ComfyUI default on smaller cards). Keeping everything loaded needs 30.9 GB.
- Diffusion model 19.0 GB
- Text encoders 7.7 GB
- VAE 0.2 GB
- Peak activations 3.1 GB
- Overhead 0.8 GB
GPU fit
- RTX 4060 8GB37% on GPU
- RTX 3060 12GB55% on GPU
- RTX 4060 Ti 16GB73% on GPU
- RTX 4090 24GBFitswith swapping
- RTX 5090 32GBFitswith swapping
- RTX 6000 Ada 48GBFitsall loaded
- RTX PRO 6000 Blackwell 96GBFitsall loaded
- Apple M4 Max 128GBFitsall loaded
The VAE decode is the peak here. Turn on tiled VAE decoding to cut it to about 0.8 GB.
Qwen-Image-Edit 2509 VRAM by precision
At 1024×1024. Text encoders in FP8 except for the FP16 row.
| Weights | Diffusion model | Peak (swapping) | All loaded | Smallest tier |
|---|---|---|---|---|
| FP16 / BF16 | 38.0 GB | 39.5 GB | 57.6 GB | 48 GB |
| FP8 | 19.0 GB | 20.5 GB | 30.9 GB | 24 GB |
| GGUF Q8_0 | 20.2 GB | 21.7 GB | 32.1 GB | 24 GB |
| GGUF Q5_K_S | 13.1 GB | 14.7 GB | 25.0 GB | 16 GB |
| GGUF Q4_K_S / NF4 | 10.8 GB | 12.4 GB | 22.7 GB | 16 GB |
VRAM by resolution
FP8 weights, with model swapping.
| Resolution | Latent tokens | Activations | Peak VRAM |
|---|---|---|---|
| 1024×1024 | 4,096 | 0.8 GB | 20.5 GB |
| 1536×1536 | 9,216 | 1.7 GB | 21.5 GB |
About Qwen-Image-Edit 2509
The editing version of Qwen-Image, with multi-image input and strong text rendering inside images. The 20B transformer is the same size as Qwen-Image, and the reference image adds extra tokens. Lightning LoRAs that cut it to 4–8 steps are widely used and do not change memory needs.
| Component | Size |
|---|---|
| Diffusion model | 20.4B parameters (MMDiT (image editing)) |
| Text encoders | Qwen2.5-VL 7B (8.3B) |
| VAE | 0.127B parameters |
| Classifier-free guidance | Yes (2 passes per step) |
| Released | 2025-09 |
| License | Apache 2.0 |
Qwen-Image-Edit 2509 VRAM: common questions
Is 16GB enough for Qwen-Image-Edit?
Yes with GGUF Q4 or Q5 weights and model swapping. FP8 weights (about 19 GB for the transformer) need a 24 GB card.
How much VRAM does Qwen-Image-Edit 2509 need?
At its default 1024×1024: about 39.5 GB in FP16 with model swapping (57.6 GB with everything loaded), 20.5 GB in FP8 and 12.4 GB with GGUF Q4 weights.