Model reference · open weights
AuraFlow is an open-weight image model from fal. AuraFlow-v0.3 (FP32) weighs 16.8 GB; the smallest configuration that runs it is RTX 4090 24 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the fal/AuraFlow-v0.3 model card, 2026-10-01
What it is
| Released by | fal |
|---|---|
| Released | 2024-08-14 |
| Parameters | 6.9B |
| VRAM | 16.8 GB for the weights |
What it runs on
| Card | One image | Memory |
|---|---|---|
| RTX 3060 12 GB … RTX 4060 Ti 16 GB | does not fit | |
| RTX 3090 24 GB | tight | 23.4 GB |
| RTX 4090 24 GB | tight | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
AuraFlow v0.3 is the fully open-sourced flow-based text-to-image generation model. The model was trained with more compute compared to the previous version, AuraFlow-v0.2.
Compared to AuraFlow-v0.2, the model is fine-tuned on more aesthetic datasets and now supports various aspect ratio, (now width and height up to 1536 pixels).
This model achieves state-of-the-art results on GenEval. Read our blog post for more technical details. You can also check out the comparison with other models on this gallery page.
The model is currently in beta. We are working on improving it and the community's feedback is important.
Credits: A huge thank you to @cloneofsimo and @isidentical for bringing this project to life. It's incredible what two cracked engineers can achieve in such a short period of time. We also extend our gratitude to the incredible researchers whose prior work laid the foundation for our efforts.
$ pip install transformers accelerate protobuf sentencepiece
$ pip install git+https://github.com/huggingface/diffusers.git
from diffusers import AuraFlowPipeline
import torch
pipeline = AuraFlowPipeline.from_pretrained(
"fal/AuraFlow-v0.3",
torch_dtype=torch.float16,
variant="fp16",
).to("cuda")
image = pipeline(
prompt="rempage of the iguana character riding F1, fast and furious, cinematic movie poster",
width=1536,
height=768,
num_inference_steps=50,
generator=torch.Generator().manual_seed(1),
guidance_scale=3.5,
).images[0]
image.save("output.png")
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour. Runs as it is with diffusers, on the machine, in Python.
# on your rented machine: pip install diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained("fal/AuraFlow-v0.3", torch_dtype=torch.bfloat16).to("cuda")
image = pipe(prompt="a red bicycle on a cobbled street").images[0]
image.save("/workspace/out.png")