Benchmark
ScreenSpot Pro
Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.
- Models measured
- 34
- Scores recorded
- 54
- Sources
- 4
- IQ weight
- 1.42
Score distribution
- 64.4%
- Mean
- 69.1%
- Median
- 22.8%
- Std dev
- 26.8%
- 10th percentile
- 87.6%
- 90th percentile
- 0.82%
- Min
This benchmark has a weight of 1.42 for chat models, measured over 34 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 92.7% |
| 2 | GPT-6 Astrano tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.7% |
| 3 | Inkling 2 Small· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 92% |
| 4 | Claude Opus 4.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.9% |
| 5 | Claude Opus 4.8Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · with tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.9% |
| 6 | Claude Opus 4.7new image resolution ≤2576px / 3.75MP · Python tools · adaptive thinking · max effort · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.6% |
| 7 | Claude Opus 4.7Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · with tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.6% |
| 8 | Claude Fable 5no tools · Mythos-derived comparison per OpenAI footnote· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.3% |
| 9 | GPT 5.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 86.3% |
| 10 | GPT 5.2Python · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 86.3% |
| 11 | Claude Opus 4.7old image resolution ≤1568px / 1.15MP · Python tools · adaptive thinking · max effort · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.9% |
| 12 | Qwen3.8 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.5% |
| 13 | Qwen3.8 MaxQwen3.8-Max release vision/video table · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.5% |
| 14 | Muse Spark· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.1% |
| 15 | Claude Opus 4.8Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · no tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.3% |
| 16 | Claude Opus 4.7new image resolution ≤2576px / 3.75MP · no tools · adaptive thinking · max effort · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 79.5% |
| 17 | Claude Opus 4.7Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · no tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 79.5% |
| 18 | Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 79% |
| 19 | Qwen3.7 PlusQwen3.7-Plus release · enable_thinking=false· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 79% |
| 20 | GPT-5.6 Solno tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 76.9% |
| 21 | Inkling 2 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.5% |
| 22 | Muse Glimmer 30B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 75.4% |
| 23 | Muse Glimmer 30Bhigh reasoning· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.4% |
| 24 | Holo2 30B A3BAgentic· ScreenSpot data | 75.2% |
| 25 | Gemini 3 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 72.7% |
| 26 | Gemini 3 Pro PreviewGemini 3 Pro Model Card · results as of November 2025· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.7% |
| 27 | Qwen3.5 122B-A10B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 70.4% |
| 28 | Qwen3.5 27B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 70.3% |
| 29 | Gemini 3 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 69.1% |
| 30 | Gemini 3 Flashno tools unless specified · media_resolution=ultra_high· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 69.1% |
| 31 | Claude Opus 4.7old image resolution ≤1568px / 1.15MP · no tools · adaptive thinking · max effort · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 69% |
| 32 | Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 68.6% |
| 33 | Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 68.2% |
| 34 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 68.2% |
| 35 | Holo2 30B A3B· ScreenSpot data | 66.1% |
| 36 | Qwen3.5 397B A17B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 65.6% |
| 37 | Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 65.6% |
| 38 | Qwen3 VL 235B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 62% |
| 39 | Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 61.8% |
| 40 | UI TARS 1.5 7B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.6% |
| 41 | qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 60.5% |
| 42 | Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 57.9% |
| 43 | qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 57.3% |
| 44 | Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 57.1% |
| 45 | Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 54.6% |
| 46 | Qwen2.5 VL 72B Instruct· ScreenSpot dataValues from 3 sources
| 53.3% |
| 47 | Qwen 2.5 VL 32B Instruct· ScreenSpot dataValues from 2 sources
| 48% |
| 48 | Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 46.6% |
| 49 | Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 29% |
| 50 | GPT 5resized, minimal· ScreenSpot data | 18.5% |
Showing 1–50 of 54 results.
Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats · ScreenSpot grounding leaderboard (GUI-Agent), Apache-2.0 (licence) · Hugging Face official benchmark leaderboards; licence varies by benchmark