FLUX.2 [klein] 9B in FP8 needs 10.6 GB of GPU memory or more (1024×1024). The cheapest option right now is RTX 3060 12 GB, from $0.084/h.
“Minimum” is what ComfyUI does by default when memory is short. “No swapping” keeps everything loaded, so repeated runs are faster.
| Weights | Model | Text encoder | Minimum | No swapping | Fits on |
|---|---|---|---|---|---|
| BF16 | 16.9 | 15.3 | 19.1 | 34.3 | RTX 4000 Ada 20 GB |
| FP8 | 8.5 | 7.6 | 10.6 | 18.3 | RTX 3060 12 GB |
| GGUF Q8 | 9.0 | 7.6 | 11.2 | 18.8 | RTX 3060 12 GB |
| GGUF Q4 | 4.7 | 7.6 | 8.1 | 14.6 | RTX 3080 10 GB |
Working memory is our estimate for the listed output size and frame count. ComfyUI also runs the model on a smaller card by swapping part of the weights to system RAM, only slower.
GPUs from the catalog the model fits on, by price per hour.
The price is the cheapest GPU of the model × the GPU count. A 2–8 GPU server can cost a little more per GPU.
The “ComfyUI” template starts the model in a ready environment: open the link once the server has loaded.
In ComfyUI the model goes into the models folder; the template ships with ComfyUI Manager.
File sizes checked against Hugging Face on 2026-10-19. 1 GB here is 1024³ bytes, like GPU memory. Hugging Face lists decimal gigabytes, so its numbers are 7% larger.