Simple text-to-image workflow for FLUX.2 klein 4B, tested on an RTX 5060 Ti 16 GB with 32 GB RAM in ComfyUI 0.38.1.
Measured: 2.9 s per 1024×1024 image, peak VRAM 12.2 GB.
Files you need:
- flux-2-klein-4b-fp8.safetensors → models/diffusion_models
- qwen_3_4b_fp8_mixed.safetensors → models/text_encoders
- flux2-vae.safetensors → models/vae
The note inside the workflow has a download link for every file.
Does it fit your card? https://vramready.com/models/flux-2-klein-4b/
All 12 tested workflows: https://vramready.com/workflows/