# GUI Automation on OSWorld Source: [GUI Automation on OSWorld]() Collection index: [llms.txt]() Data updated: 2026-07-07T10:28:09.072Z Evaluation of GUI automation performance on the OSWorld dataset, reporting success rates across specific application categories (Chrome, Multi-Apps, OS). A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | GUI automation | | Dataset (including any stated split) | OSWorld | | Metric | Overall Success Rate | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 28 of 28 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Overall Success Rate](): Higher is better; unit: PERCENT. - [Chrome Success Rate](): Higher is better; unit: PERCENT. - [Multi-Apps Success Rate](): Higher is better; unit: PERCENT. - [OS Success Rate](): Higher is better; unit: PERCENT. ## Recorded results | Method | Overall Success Rate | Result date | Report paper | | --- | --- | --- | --- | | Agentic Programming (Ours) | 86.8 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | Holo3-35B-A3B | 80.4 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | OpenAPA w/ gemini-3.1-pro | 78.3 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | VLAA-GUI w/ Opus 4.5 | 76.3 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | HIPPO Agent w/ Opus 4.5 | 74.5 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | Kimi K2.6 | 73.1 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | Agent S3 w/ Opus 4.5 + GPT-5 bBoN (N=10) | 72.6 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | claude-sonnet-4-6 | 72.1 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | Agent S3 w/ GPT-5 bBoN (N=10) | 69.9 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | UiPath Screen Agent w/ Opus 4.5 | 67.1 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | Agent S3 \[4\] w/ Opus 4.5 bBoN (N=1) | 66 | Not recorded | 6a30d8a7d33dfb23e783cd12 | | UI-MOPD | 38.2 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Qwen3-VL-235B-A22B-Thinking | 38.1 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Model Merge (TIES Merging) | 36.8 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Model Merge (Weight Averaging) | 36.5 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Mixed-SFT | 35 | Not recorded | 6a4c881e85dba69fef8b28d8 | | GUI-Owl-7B | 34.9 | Not recorded | 6a4c881e85dba69fef8b28d8 | | OpenCUA-32B | 34.8 | Not recorded | 6a4c881e85dba69fef8b28d8 | | SeedVL-1.5 | 34.1 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Qwen3-VL-8B-Instruct | 33.9 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Qwen3-VL-8B-Thinking | 33.9 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Qwen3-VL-32B-Instruct | 32.6 | Not recorded | 6a4c881e85dba69fef8b28d8 | | GELab-Zero-4B | 31.9 | Not recorded | 6a4c881e85dba69fef8b28d8 | | Qwen3-VL-235B-A22B-Instruct | 31.6 | Not recorded | 6a4c881e85dba69fef8b28d8 | | OpenAI CUA o3 | 31.3 | Not recorded | 6a4c881e85dba69fef8b28d8 | | OpenCUA-7B | 28.2 | Not recorded | 6a4c881e85dba69fef8b28d8 | | UI-TARS-1.5-7B | 27.4 | Not recorded | 6a4c881e85dba69fef8b28d8 | | UI-TARS-72B-DPO | 27.1 | Not recorded | 6a4c881e85dba69fef8b28d8 | ### Agentic Programming (Ours) Overall Success Rate: 86.8. Result ID: 6a311051fbc373592c2d867c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"15"}. This record does not assert independent reproduction. ### Holo3-35B-A3B Overall Success Rate: 80.4. Result ID: 6a311051fbc373592c2d867b. [Reporting paper]() Recorded evaluation configuration: {"Type":"Specialized model","Max Steps":"100"}. This record does not assert independent reproduction. ### OpenAPA w/ gemini-3.1-pro Overall Success Rate: 78.3. Result ID: 6a311051fbc373592c2d867a. [Reporting paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### VLAA-GUI w/ Opus 4.5 Overall Success Rate: 76.3. Result ID: 6a311051fbc373592c2d8679. [Reporting paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### HIPPO Agent w/ Opus 4.5 Overall Success Rate: 74.5. Result ID: 6a311051fbc373592c2d8678. [Reporting paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### Kimi K2.6 Overall Success Rate: 73.1. Result ID: 6a311051fbc373592c2d8677. [Reporting paper]() Recorded evaluation configuration: {"Type":"General model","Max Steps":"100"}. This record does not assert independent reproduction. ### Agent S3 w/ Opus 4.5 + GPT-5 bBoN (N=10) Overall Success Rate: 72.6. Result ID: 6a311051fbc373592c2d8676. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### claude-sonnet-4-6 Overall Success Rate: 72.1. Result ID: 6a311051fbc373592c2d8675. [Reporting paper]() Recorded evaluation configuration: {"Type":"General model","Max Steps":"100"}. This record does not assert independent reproduction. ### Agent S3 w/ GPT-5 bBoN (N=10) Overall Success Rate: 69.9. Result ID: 6a311051fbc373592c2d8674. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### UiPath Screen Agent w/ Opus 4.5 Overall Success Rate: 67.1. Result ID: 6a311051fbc373592c2d8673. [Reporting paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### Agent S3 \[4\] w/ Opus 4.5 bBoN (N=1) Overall Success Rate: 66. Result ID: 6a311051fbc373592c2d8672. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Agentic framework","Max Steps":"100"}. This record does not assert independent reproduction. ### UI-MOPD Overall Success Rate: 38.2. Result ID: 6a4cd4a460b88bb339ca92a2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Integration Strategies"}. This record does not assert independent reproduction. ### Qwen3-VL-235B-A22B-Thinking Overall Success Rate: 38.1. Result ID: 6a4cd4a460b88bb339ca9297. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### Model Merge (TIES Merging) Overall Success Rate: 36.8. Result ID: 6a4cd4a460b88bb339ca92a1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Integration Strategies"}. This record does not assert independent reproduction. ### Model Merge (Weight Averaging) Overall Success Rate: 36.5. Result ID: 6a4cd4a460b88bb339ca92a0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Integration Strategies"}. This record does not assert independent reproduction. ### Mixed-SFT Overall Success Rate: 35. Result ID: 6a4cd4a460b88bb339ca929f. [Reporting paper]() Recorded evaluation configuration: {"Category":"Integration Strategies"}. This record does not assert independent reproduction. ### GUI-Owl-7B Overall Success Rate: 34.9. Result ID: 6a4cd4a460b88bb339ca929e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"GUI Models (Multi-Platform)"}. This record does not assert independent reproduction. ### OpenCUA-32B Overall Success Rate: 34.8. Result ID: 6a4cd4a460b88bb339ca929a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"GUI Models (Single-Platform)"}. This record does not assert independent reproduction. ### SeedVL-1.5 Overall Success Rate: 34.1. Result ID: 6a4cd4a460b88bb339ca9292. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Overall Success Rate: 33.9. Result ID: 6a4cd4a460b88bb339ca9293. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Thinking Overall Success Rate: 33.9. Result ID: 6a4cd4a460b88bb339ca9294. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Instruct Overall Success Rate: 32.6. Result ID: 6a4cd4a460b88bb339ca9295. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### GELab-Zero-4B Overall Success Rate: 31.9. Result ID: 6a4cd4a460b88bb339ca929d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"GUI Models (Multi-Platform)"}. This record does not assert independent reproduction. ### Qwen3-VL-235B-A22B-Instruct Overall Success Rate: 31.6. Result ID: 6a4cd4a460b88bb339ca9296. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"General Models"}. This record does not assert independent reproduction. ### OpenAI CUA o3 Overall Success Rate: 31.3. Result ID: 6a4cd4a460b88bb339ca9299. [Reporting paper]() Recorded evaluation configuration: {"Category":"GUI Models (Single-Platform)"}. This record does not assert independent reproduction. ### OpenCUA-7B Overall Success Rate: 28.2. Result ID: 6a4cd4a460b88bb339ca9298. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"GUI Models (Single-Platform)"}. This record does not assert independent reproduction. ### UI-TARS-1.5-7B Overall Success Rate: 27.4. Result ID: 6a4cd4a460b88bb339ca929c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"GUI Models (Multi-Platform)"}. This record does not assert independent reproduction. ### UI-TARS-72B-DPO Overall Success Rate: 27.1. Result ID: 6a4cd4a460b88bb339ca929b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"GUI Models (Multi-Platform)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.