Ideogram 4

9.3B parameter DiT with Qwen3-VL text encoder, NF4 weights are about 6.5 GB and FP8 weights are about 11 GB

• BARE MINIMUM

  NVIDIA GeForce RTX 3060 12GB

  12 GB VRAM

  ~$220 USED

  NF4 quantized model, low-VRAM offload, 1024x1024, ~45 s/image

  
• SWEET SPOT

  NVIDIA GeForce RTX 4070 Ti Super 16GB

  16 GB VRAM

  ~$780 NEW

  FP8 quantized model with text encoder in VRAM, 1024x1024 to 2K, ~15 s/image

  
• FULL POWER

  NVIDIA GeForce RTX 4090 24GB

  24 GB VRAM

  ~$1800 NEW

  Full FP8 or BF16 diffusion backbone at native 2K resolution with Quality 48 steps, ~7 s/image

  

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 Ideogram 4 yet to call it. We found 188 posts but people did not say much either way.

162 POSTS · 26 COMMENTS · X · REDDIT · GITHUB