whatmodel.sh / image / DeepGen 1.0
DeepGen 1.0
↗ share
+ contribute
DeepGen 1.0 is an open-weights image model from DeepGen Team, released February 2026. Licence: apache-2.0.
Catalogue identity notes
DeepGen 1.0 unified image generation and editing model (Wang et al., 2026; deepgenteam/DeepGen-1.0, pinned b746f07, Apache-2.0; the card says it ships pre-training, SFT and RL checkpoints, the RL one being the release model). Rows: Edit-Compass 'DeepGen 1.0', ChartSync 'DeepGen-1.0' (guidance 4.0, 50 steps per its Table 7), WISE v4 Table 11 'DeepGen1.0'. 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 DeepGen 1.0, out of 197 with a score.
| # | model | score | where |
|---|---|---|---|
| 122 |  Gemini 2.0 Flash Image Generation (preview) | 45±2 | comparison only |
| 123 |  Show-o2 1.5B | 45±5 | comparison only |
| 124 | DeepGen 1.0 | 45±3 | comparison only |
| 125 |  DALL·E 3 | 44±1 | comparison only |
| 126 |  FLUX.1 Kontext [dev] | 44±1 | api $0.025/image +local |
$ what it's good at
text to image▁▂▇▇▆▅█▇▂▃▁▁ #43 of 157 * · 13% of its score · 1 test
Beats Wan 2.7 Image Pro, Wan 2.7 Image, Emu3.5-Image here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| WISE (WiScore overall, GPT-4o judge) | 73.0% ✎ | ▄·▆▂▆█▄▆▂▄·▄ beats 22/22 | 13% |
image to image / editing▁▁▁▃▃▃█▃▃▃▁▁ #66 of 81 * · 55% of its score · 1 test
Beats FLUX1.1 [pro] here, though they rank higher overall. Trails FLUX.1 Kontext [dev] here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Edit-Compass (EN, overall) | 38.2% ✎ | ▃▃▂·▃▆▂█·▃·▃ beats 8/20 | 55% |
other tests1 not in a task · 29% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| ChartSync chart editing (Overall) | 23.6% ✎ | ▃··▂█▂▄▂▂·▂▂ beats 0/12 | 29% |
$ where to run it
No verified API or download recorded yet · add one
Official weights and documentation: huggingface.co/deepgenteam ↗
$ how the score was worked out · Edit-Compass & EditReward-Com… 55%, ChartSync: A Benchmark for Vi… 29%, WISE: A World Knowledge-Infor… 13%
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 |
|---|---|---|
 Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling | 1 | 55% |
| ChartSync: A Benchmark for Visuo-Logical Cascading Chart Editing (arXiv:2607.10301v1) | 1 | 29% |
| WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation (arXiv v4) | 1 | 13% |
| prior · pull toward the middle | 3% | |
| weighted average | = 44.8 |
$ 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