whatmodel.sh / image / HiDream-O1-Image-Dev
HiDream-O1-Image-Dev
↗ share
+ contribute
HiDream-O1-Image-Dev is an open-weights image model from HiDream, released May 2026. Its published size is 8B parameters. Licence: mit.
Catalogue identity notes
HiDream's distilled Dev variant of HiDream-O1-Image (8B; 28 inference steps per the card's model table), HiDream-ai/HiDream-O1-Image-Dev pinned c0bada0, MIT, open-sourced 8 May 2026 together with the undistilled model (catalogued as hidream-o1-image). The later HiDream-O1-Image-Dev-2604 (14 May 2026) is a separate repo and is not assumed. UltraText Table 5 lists 'HiDream-O1-Image-Dev' separately from 'HiDream-O1-Image'. Comparison-only.
Model information sources
Release date ↗ · Weights licence ↗
8B as stated by HiDream for HiDream-O1-Image.
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of HiDream-O1-Image-Dev, out of 197 with a score, then more from HiDream.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 84 |  Reve Edit Fast | 55±4 | comparison only | ||
| 85 |  Qwen-Image | 54±1 | api $0.025/image +local | ||
| 86 | HiDream-O1-Image-Dev | 54±5 | comparison only | ||
| 87 |  LLaDA-Image-Turbo | 54±3 | local | ||
| 88 |  LongCat-Image | 53±4 | api $0.13/image +local | ||
| more from HiDream | |||||
| 60 | HiDream-O1-Image | 58±2 | api $0.010/image +local | ||
| 116 | HiDream-I1-Full | 47±2 | api $0.050/image +local | ||
| 190 | HiDream-E1.1 | 26±11 | comparison only | ||
$ what it's good at
text to image▁▂▇▇▆▅█▇▂▃▁▁ #67 of 157 * · 90% of its score · 1 test
Beats FLUX.2 [klein] 9B, Qwen-Image here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| UltraText Bench (Composite, bilingual dense text rendering) | 40.4% ✎ | ▁▂▂█▃▁▂▃·▁·▁ beats 6/22 | 90% |
$ where to run it
No verified API or download recorded yet · add one
Weights and documentation: huggingface.co/HiDream-ai ↗
$ how the score was worked out · UltraText Bench: A Comprehens… 90%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1) | 1 | 90% |
| prior · pull toward the middle | 10% | |
| weighted average | = 54.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json