HiDream.ai · MIT
HiDream-O1-Image
An 8B image model that works directly on pixels: generation, editing and personalised subjects. MIT licence.
How much memory it needs
Memory to run each model, from its model card or the size of its files. Smaller versions trade some quality for less memory.
| Model | Version | Needs | Business use |
|---|---|---|---|
| HiDream-O1-Image | FP8 | 8.1 GB | Allowed |
| BF16 | 16.4 GB |
Which computers run it
The best model in this group for each machine.
| Computer | Best version | Fit |
|---|---|---|
| Laptop or PC, no graphics card, 32 GB | HiDream-O1-Image FP8 | Needs a graphics card |
| RTX 4060 · 8 GB | HiDream-O1-Image FP8 | Spills into system RAM |
| RTX 3060 · 12 GB | HiDream-O1-Image FP8 | Runs well |
| RTX 5060 Ti · 16 GB | HiDream-O1-Image FP8 | Runs great |
| RTX 4090 · 24 GB | HiDream-O1-Image BF16 | Runs great |
| RTX 5090 · 32 GB | HiDream-O1-Image BF16 | Runs great |
| Mac with M4 · 16 GB | HiDream-O1-Image FP8 | Runs well |
| Mac with M4 Pro · 48 GB | HiDream-O1-Image BF16 | Runs great |
| Mac with M5 Max · 128 GB | HiDream-O1-Image BF16 | Runs great |
| Ryzen AI Max+ 395 · 128 GB | HiDream-O1-Image BF16 | Runs great |
| DGX Spark · 128 GB | HiDream-O1-Image BF16 | Runs great |
| RTX PRO 6000 · 96 GB | HiDream-O1-Image BF16 | Runs great |
| Mac Studio with M5 Ultra · 256 GB | HiDream-O1-Image BF16 | Runs great |
| Mac Studio with M3 Ultra · 512 GB | HiDream-O1-Image BF16 | Runs great |
What it is good and bad at
Good at
- One file, no separate text encoder
- Editing and subject personalisation
Not good at
- About 16 GB at full quality (8 GB at FP8)
- Low adoption so far
Start it
- HiDream-O1-Image: runs in ComfyUI (also HiDream-O1-Image repo inference.py (transformers)). Model card