Can the RTX 4070 12 GB run Qwen-Image 2.1?
Download Q8_0 (7.6 GB). With the model's working memory it needs about 9.9 GB, leaving 2.1 GB spare on 12 GB. Quality: practically identical to the original.
01What to download
| Part | File | Folder | Size | |
|---|---|---|---|---|
| Model | qwen_image_2.1_Q8_0.gguf Q8_0 · Abiray/Qwen-Image-2.1-GGUF | models/unet | 7.6 GB | Download → |
| Text encoder | qwen3vl_8b_int8_convrot.safetensors Qwen3-VL 8B INT8 · Comfy-Org/Qwen-Image-2.1 | models/text_encoders | 9.4 GB | Download → |
| VAE | qwen_image_2.1_vae_bf16.safetensors Qwen-Image 2.1 VAE · Comfy-Org/Qwen-Image-2.1 | models/vae | 0.7 GB | Download → |
| Total download · keep about the same free on disk | 17.6 GB | |||
Alternatives: Qwen3-VL 8B BF16 (17.5 GB). The 16-bit text encoder is a little more faithful; the smaller one is chosen here because it loads faster and needs less RAM. Sizes read from Hugging Face (2026-09-25). “Download” links start the file directly; the file name links are the same files the official ComfyUI workflows use.
System RAM: 32 GB or more. ComfyUI keeps the model file, the text encoder and the VAE in system RAM and moves them to the GPU as needed. With this set of files that is about 17.6 GB, plus roughly 6 GB for Windows, ComfyUI and a browser: 23.6 GB in total. With less RAM it still runs, but Windows starts swapping to disk and loading gets very slow. calculated
02Every Qwen-Image 2.1 file on 12 GB
| File | Size | Needed | On this card | Quality | Download |
|---|---|---|---|---|---|
| 16-bit SAFETENSORS · Comfy-Org | 14.2 GB | 16.5 GB | 4.5 GB too big | the original weights | Hugging Face → |
| Q8_0 ← GGUF · Abiray | 7.6 GB | 9.9 GB | fits · 2.1 GB spare | practically identical to the original | Hugging Face → |
| INT8 SAFETENSORS · Comfy-Org | 7.3 GB | 9.6 GB | fits · 2.4 GB spare | practically identical to the original | Hugging Face → |
| Q6_K GGUF · Abiray | 5.9 GB | 8.2 GB | fits · 3.8 GB spare | very close to the original | Hugging Face → |
| Q5_K_M GGUF · Abiray | 5.0 GB | 7.3 GB | fits · 4.7 GB spare | close; small differences in fine detail | Hugging Face → |
| Q4_K_M GGUF · Abiray | 4.2 GB | 6.5 GB | fits · 5.5 GB spare | good; some loss in fine detail and text | Hugging Face → |
| Q3_K_M GGUF · Abiray | 3.2 GB | 5.5 GB | fits · 6.5 GB spare | noticeable loss of detail | Hugging Face → |
Sizes from the Hugging Face file listing, read 2026-09-25. “Needed” = file + 1.5 GB working memory for this model + 0.8 GB kept free for the system.
03The text encoder
Qwen3-VL 8B: 17.5 GB as 16-bit, 9.4 GB as INT8. ComfyUI encodes the prompt first and can push the encoder out of VRAM before sampling, so it does not have to fit together with the model. On 12 GB the INT8 encoder fits on its own, so prompt encoding stays fast.
04About this GPU
The RTX 4070 12 GB is an Ada Lovelace GPU with hardware FP8, so ComfyUI can compute Comfy-Org's FP8 files natively: small and fast. (Plain FP8 files use FP8 maths with the --fast fp8_matrix_mult option.)
RTX 4070 12 GB: 12 GB GDDR6X · 192-bit · 504 GB/s · Ada Lovelace. Everything that runs on the RTX 4070 12 GB →
05What more VRAM would change
Nothing to gain for this model: the RTX 4070 12 GB already runs a top-quality file entirely in VRAM. More memory would only help with bigger images, longer clips or several models at once.
What changes from the RTX 4070 12 GB to the RTX 4070 Ti Super 16 GB → · What changes from the RTX 4070 12 GB to the RTX 3090 24 GB →
06Measured and reported results
| Label | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|
| reported | Qwen-Image-2.1 int8 convrot + qwen3vl_8b_int8 convrot TE + VAE bf16 · 1024x1024 · 25 steps “1024x1024・25ステップ（公式ワークフロー相当） | 14.0 秒” ComfyUI, RTX 4070 12GB with partial offload (17GB weights); same table: 12 steps 9.6 s, 2048x2048 20 steps 90.5 s; repo date not shown (Qwen-Image 2.1 era, 2026) | 14 s / image | — | — | github.com → |
| reported | Qwen Image 2.1 INT8 ConvRot · 832x1248 · 25 steps “「全体の所要時間」で見ると、平均で32.63秒から19.83秒へ（約39.2%短縮）… GPU全体のVRAMピーク：11.19 GiB → 11.28 GiB” ComfyUI, RTX 4070 12GB, Euler/simple CFG 1; 32.63 s = standard total time (19.83 s with Spectrum speedup node); VRAM in GiB | 32.63 s / image | 11.19 GB | 2026-09-24 | note.com → |
Reported results are other people's numbers, copied as published, with a link. Settings, drivers and ComfyUI versions differ, so compare them with care. Send yours.
07Questions
How much VRAM does Qwen-Image 2.1 need?
Around 5.5 GB with the smallest file (Q3_K_M) and 9.6 GB with an 8-bit file (INT8), counting working memory and a small system reserve. The full 16-bit file needs about 16.5 GB.
Which Qwen-Image 2.1 file should I download for the RTX 4070 12 GB?
Q8_0 (7.6 GB) from Abiray/Qwen-Image-2.1-GGUF. Load it with Unet Loader (GGUF) from the ComfyUI-GGUF node pack.
Is FP8 faster than GGUF on the RTX 4070 12 GB?
It can be. This GPU has FP8 hardware, and ComfyUI computes FP8 natively for files made for it (Comfy-Org's fp8_scaled files), or for any FP8 file with the --fast fp8_matrix_mult option. GGUF files are unpacked on the fly, which costs some speed.
How much do I need to download for Qwen-Image 2.1 on the RTX 4070 12 GB?
About 17.6 GB for the model file, text encoder and VAE listed on this page, and the same again free on disk. With a 12 GB card, 32 GB of system RAM or more is recommended.