# GUI Agent Interaction on OSWorld (Step Success Rate) Source: [GUI Agent Interaction on OSWorld (Step Success Rate)]() Collection index: [llms.txt]() Data updated: 2026-06-24T08:34:46.531Z Evaluation of GUI Agent Interaction performance on the OSWorld dataset, reporting success rates based on specific maximum step constraints (e.g., 15 steps, 50 steps). 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 Agent Interaction | | Dataset (including any stated split) | OSWorld | | Metric | Success Rate (Max Steps: 15) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 31 of 31 | | 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 - [Success Rate (Max Steps: 15)](): Higher is better; unit: PERCENT. - [All-Pass@4 Success Rate (Max Steps: 50)](): Higher is better; unit: PERCENT. - [Success Rate (Max Steps: ≥ 50)](): Higher is better; unit: PERCENT. - [Task Success Rate](): Higher is better; unit: PERCENT. ## Recorded results | Method | Success Rate (Max Steps: 15) | Result date | Report paper | | --- | --- | --- | --- | | Claude 4.5 Sonnet | 42.9 | Not recorded | 6a191d78dbd7150506f826f9 | | RoTS-32B | 42.8 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen3-VL-Plus | 33.1 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen3-VL-Flash | 32.1 | Not recorded | 6a191d78dbd7150506f826f9 | | Doubao-1.5-Thinking | 31.9 | Not recorded | 6a191d78dbd7150506f826f9 | | RoTS-7B | 31.7 | Not recorded | 6a191d78dbd7150506f826f9 | | OpenCUA-32B | 29.7 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen3-VL-8B-Thinking | 29.2 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen3-VL-32B-Thinking | 28.1 | Not recorded | 6a191d78dbd7150506f826f9 | | GUI-OWL-7B | 27.1 | Not recorded | 6a191d78dbd7150506f826f9 | | OpenAI CUA | 26 | Not recorded | 6a191d78dbd7150506f826f9 | | UI-TARS-1.5-7B | 24.5 | Not recorded | 6a191d78dbd7150506f826f9 | | OpenCUA-7B | 24.3 | Not recorded | 6a191d78dbd7150506f826f9 | | UI-TARS-72B-DPO | 24 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen2.5-VL-72B | 4.4 | Not recorded | 6a191d78dbd7150506f826f9 | | Qwen2.5-VL-32B | 3 | Not recorded | 6a191d78dbd7150506f826f9 | ### Claude 4.5 Sonnet Success Rate (Max Steps: 15): 42.9. Result ID: 6a19531afbc373592c13c7a3. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Models","Data Type":"In-house"}. This record does not assert independent reproduction. ### RoTS-32B Success Rate (Max Steps: 15): 42.8. Result ID: 6a19531afbc373592c13c7b0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"Open-source"}. This record does not assert independent reproduction. ### Qwen3-VL-Plus Success Rate (Max Steps: 15): 33.1. Result ID: 6a19531afbc373592c13c7a5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Models","Data Type":"In-house"}. This record does not assert independent reproduction. ### Qwen3-VL-Flash Success Rate (Max Steps: 15): 32.1. Result ID: 6a19531afbc373592c13c7a4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Models","Data Type":"In-house"}. This record does not assert independent reproduction. ### Doubao-1.5-Thinking Success Rate (Max Steps: 15): 31.9. Result ID: 6a19531afbc373592c13c7a2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Models","Data Type":"In-house"}. This record does not assert independent reproduction. ### RoTS-7B Success Rate (Max Steps: 15): 31.7. Result ID: 6a19531afbc373592c13c7aa. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Smaller Size)","Data Type":"Open-source"}. This record does not assert independent reproduction. ### OpenCUA-32B Success Rate (Max Steps: 15): 29.7. Result ID: 6a19531afbc373592c13c7ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"Open-source"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Thinking Success Rate (Max Steps: 15): 29.2. Result ID: 6a19531afbc373592c13c7a9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Smaller Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Thinking Success Rate (Max Steps: 15): 28.1. Result ID: 6a19531afbc373592c13c7af. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### GUI-OWL-7B Success Rate (Max Steps: 15): 27.1. Result ID: 6a19531afbc373592c13c7a8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Smaller Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### OpenAI CUA Success Rate (Max Steps: 15): 26. Result ID: 6a19531afbc373592c13c7a1. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Models","Data Type":"In-house"}. This record does not assert independent reproduction. ### UI-TARS-1.5-7B Success Rate (Max Steps: 15): 24.5. Result ID: 6a19531afbc373592c13c7a6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Smaller Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### OpenCUA-7B Success Rate (Max Steps: 15): 24.3. Result ID: 6a19531afbc373592c13c7a7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Smaller Size)","Data Type":"Open-source"}. This record does not assert independent reproduction. ### UI-TARS-72B-DPO Success Rate (Max Steps: 15): 24. Result ID: 6a19531afbc373592c13c7ad. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Success Rate (Max Steps: 15): 4.4. Result ID: 6a19531afbc373592c13c7ac. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"In-house"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B Success Rate (Max Steps: 15): 3. Result ID: 6a19531afbc373592c13c7ab. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-Weights Models (Larger Size)","Data Type":"In-house"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.