Ideogram 4 on 12 GB of VRAM
The smallest sensible file, Q4_1 (6.2 GB (×2, both loaded)), needs about 14.7 GB — 2.7 GB more than 12 GB GPU has. ComfyUI can still run it by streaming part of the model from system RAM. How much slower that is depends on your ComfyUI version, the file format and the PCIe link — see the note below.
Worth trying on NVIDIA GPUs: with an up-to-date ComfyUI, Dynamic VRAM (on by default for NVIDIA since March 2026) streams whatever does not fit from system RAM, and ComfyUI's own start-up message recommends native FP8/INT8 files over GGUF, saying they “will be faster even if they are larger than your memory”. So before settling for a heavily compressed GGUF, try the native FP8 file (9.3 GB). The file recommended above is the best one that fits entirely — the safe choice on older ComfyUI versions, AMD and Intel. Comfy blog: Dynamic VRAM →
01What to download
| Part | File | Folder | Size | |
|---|---|---|---|---|
| Model | ideogram4-transformer-q4_1.gguf main model Q4_1 · molbal/ideogram-4-gguf | models/unet | 6.2 GB | Download → |
| Model | ideogram4-unconditional_transformer-q4_1.gguf unconditional model, loaded together with the main one Q4_1 · molbal/ideogram-4-gguf | models/unet | 6.2 GB | Download → |
| Text encoder | qwen3vl_8b_fp8_scaled.safetensors Qwen3-VL 8B FP8 · Comfy-Org/Ideogram-4 | models/text_encoders | 10.6 GB | Download → |
| VAE | flux2-vae.safetensors FLUX.2 VAE · Comfy-Org/Ideogram-4 | models/vae | 0.3 GB | Download → |
| Total download · keep about the same free on disk | 23.3 GB | |||
Alternatives: Qwen3-VL 8B NVFP4 (6.3 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. FP4 files are listed only as alternatives unless the GPU is an RTX 50 (Blackwell). 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 files, 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 23.3 GB, plus roughly 6 GB for Windows, ComfyUI and a browser: 29.3 GB in total. With less RAM it still runs, but Windows starts swapping to disk and loading gets very slow. calculated
02Every file, on 12 GB
| File | Size | Needed | On this card | Quality | Download |
|---|---|---|---|---|---|
| Q8_0 GGUF · molbal | 10.1 GB ×2 | 22.6 GB | 10.6 GB too big | practically identical to the original | Hugging Face → |
| FP8 SAFETENSORS · Comfy-Org | 9.3 GB ×2 | 20.9 GB | 8.9 GB too big | practically identical to the original | Hugging Face → |
| INT8 SAFETENSORS · Comfy-Org | 9.6 GB ×2 | 21.5 GB | 9.5 GB too big | practically identical to the original | Hugging Face → |
| Q5_1 GGUF · molbal | 7.3 GB ×2 | 17.0 GB | 5.0 GB too big | close; small differences in fine detail | Hugging Face → |
| Q4_1 ← GGUF · molbal | 6.2 GB ×2 | 14.7 GB | 2.7 GB too big | good; some loss in fine detail and text | Hugging Face → |
Sizes from the Hugging Face file listing, read 2026-09-25. “Needed” = file ×2 + 1.5 GB working memory for this model + 0.8 GB kept free for the system. Needs two transformer files loaded together (main + unconditional). GGUF files need the molbal fork of ComfyUI-GGUF. No 16-bit weights were released.
03The text encoder
Qwen3-VL 8B: 10.6 GB as FP8, 6.3 GB as NVFP4. 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 FP8 encoder fits on its own, so prompt encoding stays fast.