ScreenSpot Pro

Multilingual: no

ScreenSpot-Pro is a novel GUI grounding benchmark designed to rigorously evaluate the grounding capabilities of multimodal large language models (MLLMs) in professional high-resolution computing environments. The benchmark comprises 1,581 instructions across 23 applications spanning 5 industries and 3 operating systems, featuring authentic high-resolution images from professional domains with expert annotations. Unlike previous benchmarks that focus on cropped screenshots in consumer applications, ScreenSpot-Pro addresses the complexity and diversity of real-world professional software scenarios, revealing significant performance gaps in current MLLM GUI perception capabilities.

Methodology

Imported from llm-stats public benchmark metadata. Modality: multimodal. Max score: 1. Categories: grounding, multimodal, spatial_reasoning, vision. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 20 of 22 results

Claude Opus 4.8 self-reported llm-stats

87.9%
GPT-5.2 self-reported llm-stats

86.3%
Muse Spark self-reported llm-stats

84.1%
Gemini 3 Pro self-reported llm-stats

72.7%
Qwen3.5-122B-A10B self-reported llm-stats

70.4%
Qwen3.5-27B self-reported llm-stats

70.3%
Gemini 3 Flash self-reported llm-stats

69.1%
Qwen3.5-35B-A3B self-reported llm-stats

68.6%
Qwen3.6 Plus self-reported llm-stats

68.2%
Qwen3 VL 235B A22B Instruct self-reported llm-stats

62.0%
Qwen3 VL 235B A22B Thinking self-reported llm-stats

61.8%
Qwen3 VL 30B A3B Instruct self-reported llm-stats

60.5%
Qwen3 VL 4B Instruct self-reported llm-stats

59.5%
Qwen3 VL 32B Instruct self-reported llm-stats

57.9%
Qwen3 VL 30B A3B Thinking self-reported llm-stats

57.3%
Qwen3 VL 32B Thinking self-reported llm-stats

57.1%
Qwen3 VL 8B Instruct self-reported llm-stats

54.6%
Qwen3 VL 4B Thinking self-reported llm-stats

49.2%
Qwen3 VL 8B Thinking self-reported llm-stats

46.6%
Qwen2.5 VL 72B Instruct self-reported llm-stats

43.6%