Image Gen · Model Guide · 4 min read
HiDream-I1 17B
A 17B open-weights image model that punches at the top of community quality rankings — Apache-licensed.
| Developer | HiDream.ai |
|---|---|
| Availability | Open weights |
| Commercial license | Apache 2.0 at release — commercial use permitted with attribution. Confirm the current license on the model card; HiDream has re-released variants with different terms. |
Basic Information
HiDream-I1 is a 17-billion-parameter open image model from HiDream.ai that landed near the top of open-model quality rankings at release, with the unusual bonus of an Apache 2.0 license. It runs in ComfyUI and Diffusers, and community quantizations bring it onto 16–24 GB cards.
It leans on a Mixture-of-Transconductance-free rectified-flow design with multiple text encoders, giving it strong prompt adherence and stylistic range from a single checkpoint.
Best use cases
- Local text-to-image at open-frontier quality on a 24 GB card
- Commercial image pipelines under a clean Apache 2.0 license
- Style exploration across photorealistic, anime, and illustration modes
- Community LoRA development on an open foundation
Computational Requirements
Single-GPU class: quantized community builds target 16–24 GB cards; full precision wants ~34 GB.
| Configuration | Memory |
|---|---|
| FP16 (reference) | ~32 GB |
| Quantized (community) | ~16–24 GB |
Commercial license
Apache 2.0 at release — commercial use permitted with attribution. Confirm the current license on the model card; HiDream has re-released variants with different terms.
Check whether it fits your hardware
HiDream-I1 17B is in our compatibility library — open the compatibility panel and look it up with your GPU and context length.
More guides: back to Image Gen · visit the blog