# Long-horizon web tasks on WebTailBench long-horizon tasks Source: [Long-horizon web tasks on WebTailBench long-horizon tasks]() Collection index: [llms.txt]() Data updated: 2026-05-13T07:33:36.310Z Evaluation of agent performance on long-horizon web tasks using the WebTailBench long-horizon tasks dataset, measured by success rates at different horizons (SR@50, SR@100) and efficiency metrics like average steps and tokens per step. 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 | Long-horizon web tasks | | Dataset (including any stated split) | WebTailBench long-horizon tasks | | Metric | SR@50 | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 34 of 34 | | 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 - [SR@50](): Higher is better; unit: PERCENT. - [Avg Steps@50](): Lower is better; unit: SCALAR. - [SR@100](): Higher is better; unit: PERCENT. - [Avg Steps@100](): Lower is better; unit: SCALAR. - [Avg Tokens/Step](): Lower is better; unit: SCALAR. ## Recorded results | Method | SR@50 | Result date | Report paper | | --- | --- | --- | --- | | Qwen-3-VL-32B | 40.2 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen-3-VL-32B | 39.6 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen3-VL-30B-A3B | 38 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen3-VL-30B-A3B | 37.4 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen-3-VL-32B | 34.8 | Not recorded | 6a04057bdbc836b87d82745e | | ReVision | 32.8 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen3-VL-30B-A3B | 32.5 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen-3-VL-8B | 30.4 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen-3-VL-8B | 29.8 | Not recorded | 6a04057bdbc836b87d82745e | | ReVision | 29.6 | Not recorded | 6a04057bdbc836b87d82745e | | OpenCUA | 29.5 | Not recorded | 6a04057bdbc836b87d82745e | | ReVision | 29.3 | Not recorded | 6a04057bdbc836b87d82745e | | OpenCUA | 29.1 | Not recorded | 6a04057bdbc836b87d82745e | | ReVision | 28.4 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen-3-VL-8B | 26 | Not recorded | 6a04057bdbc836b87d82745e | | OpenCUA | 25.8 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-1.5-7B-DPO | 18.7 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-1.5-7B-DPO | 18.1 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-72B-DPO | 17.4 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-72B-DPO | 16.9 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-1.5-7B-DPO | 16.5 | Not recorded | 6a04057bdbc836b87d82745e | | Kimi-VL-A3B | 15.9 | Not recorded | 6a04057bdbc836b87d82745e | | UI-TARS-72B-DPO | 15.3 | Not recorded | 6a04057bdbc836b87d82745e | | Kimi-VL-A3B | 14.8 | Not recorded | 6a04057bdbc836b87d82745e | | Kimi-VL-A3B | 13.2 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-72B | 9 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-7B | 8.6 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-72B | 8.2 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-7B | 7.9 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-32B | 7.5 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-72B | 7.5 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-32B | 6.8 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-7B | 6.5 | Not recorded | 6a04057bdbc836b87d82745e | | Qwen2.5-VL-32B | 6.2 | Not recorded | 6a04057bdbc836b87d82745e | ### Qwen-3-VL-32B SR@50: 40.2. Result ID: 6a04286a47f34ce924e9ac93. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen-3-VL-32B SR@50: 39.6. Result ID: 6a04286a47f34ce924e9ac92. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Qwen3-VL-30B-A3B SR@50: 38. Result ID: 6a04286a47f34ce924e9ac96. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen3-VL-30B-A3B SR@50: 37.4. Result ID: 6a04286a47f34ce924e9ac95. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Qwen-3-VL-32B SR@50: 34.8. Result ID: 6a04286a47f34ce924e9ac91. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### ReVision SR@50: 32.8. Result ID: 6a04286a47f34ce924e9aca3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":true}. This record does not assert independent reproduction. ### Qwen3-VL-30B-A3B SR@50: 32.5. Result ID: 6a04286a47f34ce924e9ac94. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Qwen-3-VL-8B SR@50: 30.4. Result ID: 6a04286a47f34ce924e9ac90. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen-3-VL-8B SR@50: 29.8. Result ID: 6a04286a47f34ce924e9ac8f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### ReVision SR@50: 29.6. Result ID: 6a04286a47f34ce924e9aca0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### OpenCUA SR@50: 29.5. Result ID: 6a04286a47f34ce924e9ac98. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### ReVision SR@50: 29.3. Result ID: 6a04286a47f34ce924e9aca2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### OpenCUA SR@50: 29.1. Result ID: 6a04286a47f34ce924e9ac99. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### ReVision SR@50: 28.4. Result ID: 6a04286a47f34ce924e9aca1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":true}. This record does not assert independent reproduction. ### Qwen-3-VL-8B SR@50: 26. Result ID: 6a04286a47f34ce924e9ac8e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### OpenCUA SR@50: 25.8. Result ID: 6a04286a47f34ce924e9ac97. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-1.5-7B-DPO SR@50: 18.7. Result ID: 6a04286a47f34ce924e9ac9f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-1.5-7B-DPO SR@50: 18.1. Result ID: 6a04286a47f34ce924e9ac9e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-72B-DPO SR@50: 17.4. Result ID: 6a04286a47f34ce924e9ac9c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-72B-DPO SR@50: 16.9. Result ID: 6a04286a47f34ce924e9ac9b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-1.5-7B-DPO SR@50: 16.5. Result ID: 6a04286a47f34ce924e9ac9d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Kimi-VL-A3B SR@50: 15.9. Result ID: 6a04286a47f34ce924e9ac8d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### UI-TARS-72B-DPO SR@50: 15.3. Result ID: 6a04286a47f34ce924e9ac9a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Kimi-VL-A3B SR@50: 14.8. Result ID: 6a04286a47f34ce924e9ac8c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Kimi-VL-A3B SR@50: 13.2. Result ID: 6a04286a47f34ce924e9ac8b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B SR@50: 9. Result ID: 6a04286a47f34ce924e9ac8a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B SR@50: 8.6. Result ID: 6a04286a47f34ce924e9ac84. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B SR@50: 8.2. Result ID: 6a04286a47f34ce924e9ac89. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B SR@50: 7.9. Result ID: 6a04286a47f34ce924e9ac83. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B SR@50: 7.5. Result ID: 6a04286a47f34ce924e9ac87. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":5,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B SR@50: 7.5. Result ID: 6a04286a47f34ce924e9ac88. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B SR@50: 6.8. Result ID: 6a04286a47f34ce924e9ac86. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":3,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B SR@50: 6.5. Result ID: 6a04286a47f34ce924e9ac82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B SR@50: 6.2. Result ID: 6a04286a47f34ce924e9ac85. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"History":1,"Drop":false}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.