Ming-Image 0.1 Design VRAM requirements

01The files, and how much VRAM each needs

“Needed” = file + 1.5 GB working memory + 0.8 GB system reserve. Made for 2048×2048 as well as 1024×1024; at 2048 the working memory is several times larger than the figure used here. A separate "layer" checkpoint of the same size splits a design into layers. Native in ComfyUI since v0.38.0 (late September 2026).

02By amount of VRAM

03Best GPU for Ming-Image

The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for Ming-Image → Planning bigger images or longer clips? Open the calculator →

04By graphics card

On RTX 40/50 GPUs the FP8 file is preferred over Q8_0 when both fit (hardware FP8). All verdicts are calculated; see how the numbers work.

05Text encoder, VAE and other files

The files the official ComfyUI workflows load next to the model. Sizes read from Hugging Face (2026-10-01). Every GPU page for this model lists the exact set to download for that card, with the total.

Ling-Mini-2.0 (a large language model): 36.7 GB as 16-bit, 19.5 GB as INT8, 12.8 GB as 4-bit w4a8 (GGUF versions from realrebelai: Q4_K_M 11.5 GB, Q2_K 7.9 GB). 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. Use the smallest encoder file; it still fits on its own.

06Where the files go in ComfyUI

Standard ComfyUI folders. After copying files, press R in ComfyUI (or restart it) to refresh the lists. Some uploads need their uploader's own loader node — see the notes above.

07AMD, Intel and NVIDIA: which file types are fast

Every file type loads on every listed GPU, so the memory verdicts apply to all of them. What differs is speed: FP8 maths needs RTX 40/50 or RX 9000 (with ROCm 6.4+ and PyTorch 2.7+); ComfyUI's INT8 maths runs on NVIDIA and AMD, not on Intel; GGUF is unpacked on the fly on any GPU, which costs some speed. NVFP4 files are fast only on RTX 50. AMD runs ComfyUI on Windows through ROCm, Intel through PyTorch XPU; some custom nodes are NVIDIA-only. Source: ComfyUI model_management.py. AMD and Intel guide →

08Measured and reported results

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.

09Training a LoRA for Ming-Image

No trainer documentation with a VRAM figure for this model was found yet. What the trainers say for other models →

10Every file tracked for Ming-Image

6B DiT (30 layers) for text-heavy design: UI, posters, infographics; native RGBA/transparent output; 1024 and 2048 native res (2048 recommended); 12 steps, CFG 1.0. A separate 'layer' checkpoint (same size) does layer decomposition. The text encoder is very large (Ling-Mini-2.0 MoE-based, 36.7 GB in bf16), which dominates memory needs; the official repo validated on one 80 GiB GPU. Comfy-Org also hosts a "layer" checkpoint (same size) for splitting a design into layers.