whatmodel.sh / image / UltraEdit (SD3, with mask)
UltraEdit (SD3, with mask)
↗ share
+ contribute
UltraEdit (SD3, with mask) is an open-weights image model from Peking University / BIGAI (UltraEdit), released June 2024.
Catalogue identity notes
UltraEdit (Zhao et al., NeurIPS 2024 D&B) SD3-based editing checkpoint BleachNick/SD3_UltraEdit_w_mask (pinned e5beecb, created 2024-06-30). EdiVal: 'UltraEdit: settings from https://github.com/HaozheZhao/UltraEdit; we apply a black mask since no explicit mask is provided', i.e. the mask-conditioned SD3 checkpoint used mask-free (the repo README's free-form recipe). Licence: the card says 'More Information Needed', so licence is left unset. WEAVE 'UltraEdit(SD3)' and Edit-Compass 'UltraEdit' do not say which SD3 checkpoint (w_mask vs freeform) and are held. Comparison-only.
Model information sources
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of UltraEdit (SD3, with mask), out of 197 with a score.
| # | model | score | where |
|---|---|---|---|
| 171 |  IF-L | 33±3 | local |
| 172 |  Lumina-DiMOO | 32±4 | local |
| 173 | UltraEdit (SD3, with mask) | 32±7 | comparison only |
| 174 |  VILA-U 7B (256px) | 32±7 | local |
| 175 |  LCM-SSD-1B | 32±4 | local |
$ what it's good at
other tests1 not in a task · 80% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| EdiVal multi-turn editing overall (EdiVal-O after turn 3) | 22.9% ✎ | ▂·▂▃·▂·▃▂·▃█ beats 3/14 | 80% |
$ where to run it
No verified API or download recorded yet · add one
Weights and documentation: huggingface.co/BleachNick ↗
$ how the score was worked out · EdiVal-Agent: An Object-Centr… 80%
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 |
|---|---|---|
 EdiVal-Agent: An Object-Centric Framework for Automated, Fine-Grained Evaluation of Multi-Turn Editing (arXiv:2509.13399v3) | 1 | 80% |
| prior · pull toward the middle | 20% | |
| weighted average | = 32.4 |
$ 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