Flux 2
32 billion parameter flow matching transformer by the creators of Stable Diffusion — with open weights and LoRA fine-tuning support.
Resolution
1024x1024, Up to 4MP (2048x2048)
Aspect Ratios
1:1, 3:2, 2:3, 16:9, 9:16
What makes Flux 2 unique
32 billion parameter latent flow matching transformer — fundamentally different from diffusion models, enabling fewer sampling steps and better prompt adherence
Built by Black Forest Labs — the team behind Stable Diffusion
Open-weight Dev variant available for fine-tuning and self-hosting
LoRA-compatible for brand-consistent fine-tuning
Processes up to 10 reference images simultaneously
Four model tiers: Klein (real-time, <1s), Dev, Pro, and Max (highest quality)
Best for
How Flux 2 works
Flux 2 uses a rectified flow matching architecture coupled with a Mistral-3 24B vision-language model. Unlike traditional diffusion, flow matching learns direct mappings between text and image representations, resulting in fewer sampling steps and superior prompt adherence. The 32B parameter model achieves ~60% first-attempt accuracy on complex typography.
How to use Flux 2 on Flashloop
Select Flux 2
Open Flashloop and choose Flux 2 from the model selector in the image creator.
Enter your prompt
Describe what you want to generate — be as detailed as you like.
Generate & download
Hit generate and your image will be ready in seconds. Download or share directly.
Ready to create with Flux 2?
Start generating images with Flux 2 on Flashloop — available on iOS, Android, and web.
Start Creating with Flux 2