whatmodel.sh / image / MagicBrush (InstructPix2Pix fine-tune, jul7)
MagicBrush (InstructPix2Pix fine-tune, jul7)
↗ share
+ contribute
MagicBrush (InstructPix2Pix fine-tune, jul7) is an open-weights image model from Ohio State University (OSU NLP), released July 2023. Licence: creativeml-openrail-m.
Catalogue identity notes
MagicBrush (Zhang et al., NeurIPS 2023 Datasets & Benchmarks): InstructPix2Pix fine-tuned on the MagicBrush dataset. EdiVal Appendix pins 'Magicbrush: same settings as IP2P; model weights from https://huggingface.co/vinesmsuic/magicbrush-jul7' (pinned d75a9f1, created 2023-07-08; card licence creativeml-openrail-m). Only the EdiVal row is mapped; Edit-Compass, UnicBench, WEAVE and ImagenHub print 'MagicBrush' without naming the checkpoint (the same uploader also has magicbrush-paper), so those rows are held. Comparison-only: no local command verified.
Model information sources
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of MagicBrush (InstructPix2Pix fine-tune, jul7), out of 197 with a score.
| # | model | score | where |
|---|---|---|---|
| 182 |  Versatile Diffusion v1.0 | 29±4 | local |
| 183 |  DALL·E Mega (minDALL-E) | 29±4 | local |
| 184 | MagicBrush (InstructPix2Pix fine-tune, jul7) | 29±7 | comparison only |
| 185 |  Openjourney | 28±2 | local |
| 186 |  Stable Diffusion v1.4 | 27±2 | local |
$ what it's good at
other tests1 not in a task · 78% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| EdiVal multi-turn editing overall (EdiVal-O after turn 3) | 19.4% ✎ | ▂·▃▃·▂·▃▂·▃█ beats 1/14 | 78% |
$ where to run it
No verified API or download recorded yet · add one
Weights and documentation: huggingface.co/vinesmsuic ↗
$ how the score was worked out · EdiVal-Agent: An Object-Centr… 78%
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 | 78% |
| prior · pull toward the middle | 22% | |
| weighted average | = 28.7 |
$ 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