Qwen-Image-Edit-2511
Qwen-Image-Edit-2511 needs 20.8 GB of VRAM in ComfyUI, at the diffusion stage of its own template, by file sizes read 3 November 2026. The smallest card that loads it with room to work is the RTX 5090 (32 GB), and the template downloads 30.2 GB.
- Made by
- Qwen (Alibaba)
- Licence
- Apache 2.0
- Released
- December 2025
- Sizes read
- 3 November 2026
Largest stage from exact file sizes, with ComfyUI's own template; LoRAs are not counted. Working memory for the generation comes on top. Open it in the ComfyUI checker
Which file on your card
For each memory size, the best file for that card: the highest-precision diffusion file that loads with room to work (one that only just fits when nothing else does), and what is left for working memory while it samples. On a larger card that can be above the file ComfyUI's template loads, which the answer above is about. "Streams" means even the smallest file is larger than the card: ComfyUI still runs it from system RAM, more slowly. FP4 files go only to the Blackwell cards that run them natively.
ComfyUI's template: FP8 mixed diffusion model, FP8 scaled text encoder. Stages: text encoder 9.4 GB, diffusion 20.8 GB. Download 30.2 GB.
Computed 3 November 2026Check another card, or pick a file by hand, in the ComfyUI VRAM checker. Built your own workflow around it, with LoRAs or custom nodes? Drop it into the ComfyUI workflow checker.
Where to run it
ComfyUI's Qwen-Image-Edit-2511 workflow peaks at 20.8 GB, more than a 24 GB card holds with room to work; a 48 GB card holds it. Every way to run ComfyUI in the cloud, compared.
On-demand prices before tax, read 19 October 2026.
About the model
Qwen-Image-Edit-2511 is Qwen's open image-editing model. Qwen's image-editing model from December 2025: give it an image and an instruction. A 20-billion-parameter model under Apache 2.0, with the same Qwen2.5-VL text encoder as the original Qwen-Image.
20.4B parameters Text encoder: Qwen2.5-VL 7B Apache 2.0 Released December 2025
Good at
- Instruction-based edits: material swaps, object changes, restyling.
- Commercial use under Apache 2.0.
- LoRA training on cards from 12 GB up, with musubi-tuner's documented settings.
Watch out for
- ComfyUI's template loads the FP8 mixed model with a 4-step Lightning LoRA, not the BF16 original, which is twice the size.
- Qwen's "within 4GB VRAM" figure is for DiffSynth-Studio's layer-by-layer offloading, not ComfyUI.
Every Qwen-Image-Edit-2511 file ComfyUI loads
Sizes are the exact byte counts Hugging Face reports, read 3 November 2026. Files marked "template" are the ones ComfyUI's official workflow downloads.
| Part | Precision | Size | File |
|---|---|---|---|
| Diffusion model | BF16 | 40.86 GB | qwen_image_edit_2511_bf16.safetensors |
| Diffusion model | FP8 mixed · template | 20.53 GB | qwen_image_edit_2511_fp8mixed.safetensors |
| Diffusion model | INT8 ConvRot | 20.50 GB | qwen_image_edit_2511_int8_convrot.safetensors |
| Text encoder | BF16 | 16.58 GB | qwen_2.5_vl_7b.safetensors |
| Text encoder | FP8 scaled · template | 9.38 GB | qwen_2.5_vl_7b_fp8_scaled.safetensors |
| Text encoder | NVFP4 | 6.11 GB | qwen_2.5_vl_7b_nvfp4.safetensors |
| VAE | BF16 · template | 0.25 GB | qwen_image_vae.safetensors |
What Qwen says it needs
- "low-GPU-memory layer-by-layer offload (inference within 4GB VRAM)" (DiffSynth-Studio's layer-by-layer offloading, not ComfyUI, source)
Those figures and this page measure different things. A vendor's figure covers its own script at its default resolution, working memory included. This page counts the weights each stage holds, which is exact, and leaves the working memory as the space that remains, because it depends on your resolution and attention kernel and nobody publishes it per model. If generation runs out of memory with a file that loads, lower the resolution before stepping down a precision. ComfyUI's README says it "can run even the biggest open source models on as low as 4GB vram + 8GB ram" by streaming weights (source), so "streams" means slower, not impossible.
LoRA training for Qwen-Image-Edit-2511
Only configurations a trainer's own documentation gives a VRAM figure for. Settings change the figure a great deal, so read the setting column before trusting the number.
| Trainer | VRAM | Setting | Source |
|---|---|---|---|
| musubi-tuner | 42 GB | 1024×1024, batch 1, no quantisation; measured on Qwen-Image, and Edit needs more for its control images | docs |
| musubi-tuner | 24 GB | as above with an FP8 model and 16 blocks swapped | docs |
| musubi-tuner | 12 GB | as above with 45 blocks swapped (64 GB of system RAM recommended) | docs |
Questions
- How much VRAM does Qwen-Image-Edit-2511 need?
- In ComfyUI's own workflow, the largest stage is the diffusion stage at 20.8 GB, and the whole workflow downloads 30.2 GB. Working memory for generation comes on top and grows with resolution. Qwen itself says: "low-GPU-memory layer-by-layer offload (inference within 4GB VRAM)" (DiffSynth-Studio's layer-by-layer offloading, not ComfyUI).
- Can I run Qwen-Image-Edit-2511 on 8 GB of VRAM?
- Not fully. The smallest file here, FP8 mixed at 20.5 GB, needs 20.8 GB at its largest stage, more than 8 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.
- Can I run Qwen-Image-Edit-2511 on 12 GB of VRAM?
- Not fully. The smallest file here, FP8 mixed at 20.5 GB, needs 20.8 GB at its largest stage, more than 12 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.
- Can I run Qwen-Image-Edit-2511 on 16 GB of VRAM?
- Not fully. The smallest file here, FP8 mixed at 20.5 GB, needs 20.8 GB at its largest stage, more than 16 GB holds. ComfyUI will still run it by streaming weights from system RAM, which works but is slower.
- Which Qwen-Image-Edit-2511 file should I download?
- The highest precision that loads fully on your card. On a 24 GB card that is the FP8 mixed file (20.5 GB). ComfyUI's template downloads FP8 mixed with the FP8 scaled text encoder. FP4 files (NVFP4) are made for Blackwell cards (RTX 50, RTX PRO 6000, DGX Spark); on older cards, use FP8 or INT8.
- Files and sizes: Comfy-Org/Qwen-Image-Edit_ComfyUI on Hugging Face, read from the Hugging Face API. Template files: ComfyUI's Qwen-Image-Edit-2511 guide.
- Model and licence: Qwen/Qwen-Image-Edit-2511 and the Apache 2.0.
- Card memory: manufacturers' specifications, as on each GPU page. The checker covers NVIDIA cards; ComfyUI also runs on AMD and Apple silicon, where file formats and speed differ and we have not sized them.
- Fit rule: a stage loads fully under 85% of usable memory and is tight up to 95%, the same margins the LLM pages use.
- Prices: each provider's own pricing page, on the dates beside them.