Model CoverageDiffusionBlack Forest LabsFLUX.2-dev

FLUX.2-dev

View as Markdown

FLUX.2-dev is Black Forest Labs’ 32B rectified flow transformer for text-to-image generation and single- or multi-reference image editing.

NeMo AutoModel includes full-parameter and Low-Rank Adaptation (LoRA) text-to-image fine-tuning recipes for this checkpoint.

Fine-Tune FLUX.2-dev

From the repository root, run:

uv run torchrun --nproc-per-node=8 examples/diffusion/finetune/finetune.py \
  -c examples/diffusion/finetune/flux2_t2i_flow.yaml

Model Reference

Model Architecture

PropertyValue
Upstream TasksText-to-image generation; single- and multi-reference image editing
Model TypeRectified flow transformer
Parameters32B
Reference ArchitectureFlux2
Transformer Blocks8 double-stream / 48 single-stream
Hidden Size6,144
Attention48 heads
Feed-Forward Ratio3.0
Text EncoderMistral-Small-3.2-24B-Instruct-2506
Text Feature Width15,360
Latent Channels128 input / 128 output
Position Axes4 axes with 32 dimensions each
Rotary Position Embedding Theta2,000
GuidanceGuidance-distilled with guidance embedding enabled
Weight DtypeBF16

Example Model and Recipes

See the Diffusion Fine-Tuning Guide.