by fal
Open-source DiT model from fal.ai combining MMDiT and single DiT blocks in a Flux-like hybrid architecture. 6.35B transformer with T5-XL (~1.2B) text encoder. Apache 2.0 licensed — fully open for commercial use.
VRAM requirements, GPU fit, and setup notes for AuraFlow v0.3, including 8GB/12GB fit guidance where relevant. Recommended runtimes: ComfyUI and Diffusers support. Best download size: ~12.7 GB at FP16.
Your hardware
Detecting...
Measured quality metrics for AuraFlow v0.3 outputs.
How often humans prefer this model's output (0-100%)
Visual quality and composition rating (5-9 scale)
Text-image alignment accuracy (higher is better)
Compare which GPUs can run AuraFlow v0.3 at different precisions. FP8 uses less memory than FP16 when available, and the grade shows how comfortably each GPU handles the workload.
| Resolution | VRAM Required | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 | 16.5 GB | S● | D● | F● | B● |
| 768×768 | 16.7 GB | S● | D● | F● | B● |
| 1024×1024 | 17.0 GB | S● | F● | F● | B● |
Turbo / LCM distillation
Use distilled scheduler at 4-8 steps for faster iteration
from diffusers import AuraFlowPipeline
import torch
pipe = AuraFlowPipeline.from_pretrained(
"fal/AuraFlow-v0.3",
torch_dtype=torch.float16
)
pipe.to("cuda")
image = pipe(
prompt="your prompt here",
num_inference_steps=50,
guidance_scale=3.5,
height=1536,
width=1536,
).images[0]
image.save("output.png")Get started
Setup instructions for running AuraFlow v0.3 locally
1. Download the model
Get the checkpoint from HuggingFace
2. Place in:
ComfyUI/models/checkpoints/3. Launch ComfyUI
python main.pyVRAM allocation at 1024×1024 on RTX 4090 24GB (24 GB)
Time per image at 1024×1024, 28 steps, FP16.
Download AuraFlow v0.3 in the precision that matches your GPU. Lower precision usually means less VRAM pressure, while higher precision keeps more quality.
| Format | Precision | Size | Provider | |
|---|---|---|---|---|
| safetensorsRecommended | FP16 | 12.7 GB | official | Download |
Frequently asked questions
AuraFlow v0.3 (6.35B parameters) requires approximately 17.0 GB of VRAM at FP16 precision for standard 1024×1024 image generation. If you want a lighter setup, lower precisions like FP8 can reduce memory use when available.
AuraFlow v0.3 usually needs more than 8GB for comfortable local use. Check the VRAM table above for the exact resolution and precision trade-off.
AuraFlow v0.3 is marked for ComfyUI and Diffusers support in our catalog, so those are the runtimes we recommend first for local setup. If your workflow uses another front end, check the model's available formats and workflow notes above before downloading.
Yes, the RTX 4090 (24 GB VRAM) can run AuraFlow v0.3 comfortably at FP16. Expected generation time is around ~18s per image at 1024×1024.
There are currently no known ControlNet adapters for AuraFlow v0.3. Check Hugging Face and Civitai for community-contributed adapters.
No LoRA ecosystem. AuraFlow is relatively niche compared to Flux and SDXL. The LoRA ecosystem for AuraFlow v0.3 is rated as "none". Each LoRA adds roughly 0.0 GB of extra VRAM.
On a reference GPU (RTX 4090 24GB), AuraFlow v0.3 generates a 1024×1024 image in approximately ~18s at FP16 with 28 inference steps. Faster GPUs with higher memory bandwidth will produce images more quickly.
See also