OmniParser
A simple screen parsing tool towards pure vision based GUI agent
OmniParser scores 5.2/10 (Solid) in Browser & Computer Use.
Language Jupyter Notebook License CC-BY-4.0 Stars 25k★ Type model Self-host Yes Repo microsoft/OmniParser Homepage https://microsoft.github.io/OmniParser/
Quality
Solid 5.2
Adoption 60
Activity 17
Maturity 87
Community 59
Capability 40
Show the math
Overall 5.2/10
value = 5.2/10
score = value
Adoption 6.4/10
value = 59.9/100
score = 1 + 9 * value / 100
Activity 2.5/10
value = 17/100
score = 1 + 9 * value / 100
Maturity 8.8/10
value = 86.7/100
score = 1 + 9 * value / 100
Community 6.3/10
value = 59.4/100
score = 1 + 9 * value / 100
Capability 4.6/10
value = 39.6/100
score = 1 + 9 * value / 100
Key metrics
25k★ GitHub stars
0/90d Recent commits
0 devs/90d Recent contributors
ScreenSpot Pro 39.6% ScreenSpot Pro GUI grounding accuracy (OmniParser V2 + GPT-4o) OmniParser V2 + GPT-4o reaches state-of-the-art 39.6 average accuracy on the ScreenSpot Pro grounding benchmark. Not a WebArena/OSWorld/WebVoyager task-success number but the standard published capability metric for this UI-grounding tool. Source: Microsoft Research, https://www.microsoft.com/en-us/research/articles/omniparser-v2-turning-any-llm-into-a-computer-use-agent/
1.8y Project age
3mo ago Last commit
Gotchas
No gotchas documented yet. Contribute one if you know a constraint we missed.
Labels
- Self-hosted
Self-hosted
You deploy and operate it yourself; there is no hosted option here.
grounding vision screen-parsing gui perception