FLUX.1 Dev
#52 in Text-to-Imageblack-forest-labs · since 2024-08-01 · 2× · last seen Jun 30, 2026
1
Momentum
FLUX.1 [dev] is an open-weight, guidance-distilled text-to-image model by Black Forest Labs with 12 billion parameters, based on a rectified flow transformer architecture. It was directly distilled from FLUX.1 [pro] and achieves comparable quality and prompt adherence at improved efficiency. Weights are available on HuggingFace under a non-commercial license; commercial use requires a separate agreement with Black Forest Labs. The model is intended for research, artistic workflows, and non-commercial applications.
Momentum trend
04.04.03.07.
Features
| API Availability | Yes – weights on HuggingFace; inference API via BFL (docs.bfl.ai), Replicate, fal.ai, DeepInfra, mystic, tensorOpera, hyperbolic, and NVIDIA NIM (build.nvidia.com) |
| Image Resolution (Max.) | Natively trained for 1024×1024 px; via NVIDIA NIM API supported outputs include 1344×768, 1216×832 px (additional aspect ratios possible) |
| Generation Speed | ~15–30 seconds per image (1024×1024, 20 steps, FP16, RTX 4090); recommended step count: 20–50 steps; on RTX 3090 (24 GB FP16) approx. 10–18 s at 20 steps |