BlogLLM Benchmark Scores as Published, Side by Side
ScreenSpot-v2: Results as Published
Every result the catalog holds for ScreenSpot-v2: 9 models from 4 developers, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Computer UseMaintainer OS-Copilot (Zhiyong Wu and co-authors)Introduced 2024Data version 2026103001, as of 2026-10-30
ScreenSpot-v2 is a corrected edition of the ScreenSpot grounding test for graphical interfaces. The model sees a mobile, desktop, or web screenshot and an instruction, and it has to locate the element the instruction names. The authors replaced bad or ambiguous items and corrected wrong boxes, and they kept the same number of test items. A prediction is correct when the predicted location falls inside the ground-truth box, and the reported score is that grounding accuracy.
ScreenSpot-v2
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Gemma 3 12BGoogle, released 2025-03-12 | 3.5% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit |
| Qwen3-VL-32B-InstructAlibaba, released 2025-10-21 | 93.7% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit; Context: 4096 |
| 93.9% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit; Context: 32768 | |
| Qwen3-VL-32B-ThinkingAlibaba, released 2025-10-21 | 83.1% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: thinking; Harness: VLMEvalKit; Context: 4096 |
| 83.1% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: thinking; Harness: VLMEvalKit; Context: 40960 | |
| Qwen3-VL-8B-InstructAlibaba, released 2025-10-15 | 91.5% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit; Context: 4096 |
| 91.5% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit; Context: 32768 | |
| Qwen3-VL-8B-ThinkingAlibaba, released 2025-10-15 | 93.3% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: thinking; Harness: VLMEvalKit; Context: 4096 |
| 92.7% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: thinking; Harness: VLMEvalKit; Context: 40960 | |
| Kimi-VL-A3B-InstructMoonshot AI, released 2025-04-10 | 89.8% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit |
| Kimi-VL-A3B-ThinkingMoonshot AI, released 2025-04-10 | 81.8% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: thinking; Harness: VLMEvalKit |
| Phi-4-reasoning-vision-15BMicrosoft, released 2026-03-04 | 88.3% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: force nothink; Harness: VLMEvalKit |
| 88.1% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Effort: force thinking; Harness: VLMEvalKit | |
| 88.2% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit | |
| Phi-4-multimodal-instructMicrosoft, released 2025-02-26 | 28.5% | Microsoft, Microsoft Phi-4-reasoning-vision-15B model cardchecked 2026-10-15 | Harness: VLMEvalKit |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.