Qwen Image Lightning

20B base MMDiT parameters with distilled 4/8-step Lightning LoRA adapters, FP8 base model is about 20 GB

• BARE MINIMUM

  NVIDIA GeForce RTX 3060 12GB

  12 GB VRAM (PLUS 32 GB SYSTEM RAM)

  ~$260 USED

  GGUF Q4 or FP8 with CPU offloading via ComfyUI, 4-step generation

  
• SWEET SPOT

  NVIDIA GeForce RTX 4070 Ti Super 16GB

  16 GB VRAM (PLUS 32 GB SYSTEM RAM)

  ~$750 USED

  FP8 unet and text encoder in VRAM, 4-step Lightning generation

  
• FULL POWER

  NVIDIA GeForce RTX 4090 24GB

  24 GB VRAM

  ~$1,750 USED

  BF16/FP8 full precision pipeline, 8-step Lightning generation

  

HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.

STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.

Not enough discussion about Qwen Image Lightning yet to call it. We found 37 posts but people did not say much either way.

34 POSTS · 3 COMMENTS · DEV FORUMS · GITHUB · LEMMY · REDDIT