Datasheet for local AI49 models98 GPUsData read 2026-09-25
No adsNo tracking
Check my GPU →

Ideogram 4 on 24 GB of VRAM

Image model · 9.3BAny 24 GB GPUData 2026-09-25
Runs well
Yes — and comfortably.

Download Q8_0 (10.1 GB (×2, both loaded)). With the model's working memory it needs about 22.6 GB, leaving 1.4 GB spare on 24 GB. Quality: practically identical to the original.

Best fileQ8_0
File size10.1 GB ×2
VRAM needed~22.6 GB
System RAM48 GB+
Memory map · 24 GB GPU22.6 GB / 24 GB
012 GB24 GB
Weights Q8_0 ×2 · 20.3 GBWorking memory · 1.5 GBReserve · 0.8 GBFree · 1.4 GB
calculated from real file sizes plus working memory. Not yet measured on this setup. How this works.

On RTX 40 and RTX 50 GPUs take the FP8 file instead of Q8_0: same class, faster there. On older NVIDIA, AMD and Intel GPUs Q8_0 is the better 8-bit file.

01What to download

PartFileFolderSize
Modelideogram4-transformer-q8_0.gguf
main model
Q8_0 · molbal/ideogram-4-gguf
models/unet10.1 GBDownload →
Modelideogram4-unconditional_transformer-q8_0.gguf
unconditional model, loaded together with the main one
Q8_0 · molbal/ideogram-4-gguf
models/unet10.1 GBDownload →
Text encoderqwen3vl_8b_fp8_scaled.safetensors
Qwen3-VL 8B FP8 · Comfy-Org/Ideogram-4
models/text_encoders10.6 GBDownload →
VAEflux2-vae.safetensors
FLUX.2 VAE · Comfy-Org/Ideogram-4
models/vae0.3 GBDownload →
Total download · keep about the same free on disk31.2 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: 48 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 31.2 GB, plus roughly 6 GB for Windows, ComfyUI and a browser: 37.2 GB in total. With less RAM it still runs, but Windows starts swapping to disk and loading gets very slow. calculated

02Every file, on 24 GB

FileSizeNeededOn this cardQualityDownload
Q8_0 ←
GGUF · molbal
10.1 GB ×222.6 GBfits · 1.4 GB sparepractically identical to the originalHugging Face →
FP8
SAFETENSORS · Comfy-Org
9.3 GB ×220.9 GBfits · 3.1 GB sparepractically identical to the originalHugging Face →
INT8
SAFETENSORS · Comfy-Org
9.6 GB ×221.5 GBfits · 2.5 GB sparepractically identical to the originalHugging Face →
Q5_1
GGUF · molbal
7.3 GB ×217.0 GBfits · 7.0 GB spareclose; small differences in fine detailHugging Face →
Q4_1
GGUF · molbal
6.2 GB ×214.7 GBfits · 9.3 GB sparegood; some loss in fine detail and textHugging 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 24 GB the FP8 encoder fits on its own, so prompt encoding stays fast.