multimodal benchmark
ScreenSpot Pro is a multimodal GUI grounding benchmark for evaluating MLLM performance in professional high-resolution computing environments.
Updated Sep 5, 2026
Higher score ranks better on this benchmark.
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | Score92.70% | Percentile100.00% | Participants26 | EvidenceC | Evaluated |
| Rank02 | ModelAN | Score87.90% | Percentile96.00% | Participants26 | EvidenceC | Evaluated |
| Rank03 | ModelOP | Score86.30% | Percentile92.00% | Participants26 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score84.50% | Percentile88.00% | Participants26 | EvidenceC | Evaluated |
| Rank05 | ModelME | Score84.10% | Percentile84.00% | Participants26 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score79.00% | Percentile80.00% | Participants26 | EvidenceC | Evaluated |
| Rank07 | ModelME | Score75.40% | Percentile76.00% | Participants26 | EvidenceC | Evaluated |
| Rank08 | ModelGO | Score72.70% | Percentile72.00% | Participants26 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score70.40% | Percentile68.00% | Participants26 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score70.30% | Percentile64.00% | Participants26 | EvidenceC | Evaluated |
| Rank11 | ModelGO | Score69.10% | Percentile60.00% | Participants26 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score68.60% | Percentile56.00% | Participants26 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score68.20% | Percentile52.00% | Participants26 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score62.00% | Percentile48.00% | Participants26 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score61.80% | Percentile44.00% | Participants26 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score60.50% | Percentile40.00% | Participants26 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score59.50% | Percentile36.00% | Participants26 | EvidenceC | Evaluated |
| Rank18 | ModelAC | Score57.90% | Percentile32.00% | Participants26 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score57.30% | Percentile28.00% | Participants26 | EvidenceC | Evaluated |
| Rank20 | ModelAC | Score57.10% | Percentile24.00% | Participants26 | EvidenceC | Evaluated |
| Rank21 | ModelAC | Score54.60% | Percentile20.00% | Participants26 | EvidenceC | Evaluated |
| Rank22 | ModelAC | Score49.20% | Percentile16.00% | Participants26 | EvidenceC | Evaluated |
| Rank23 | ModelAC | Score46.60% | Percentile12.00% | Participants26 | EvidenceC | Evaluated |
| Rank24 | ModelAC | Score43.60% | Percentile8.00% | Participants26 | EvidenceC | Evaluated |
| Rank25 | ModelAC | Score39.40% | Percentile4.00% | Participants26 | EvidenceC | Evaluated |
| Rank26 | ModelAC | Score29.00% | Percentile0.00% | Participants26 | EvidenceC | Evaluated |
The leading models and scores on this benchmark.
A closer view of the leading scores on this benchmark.
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What ScreenSpot Pro measures and how its scores work.
ScreenSpot Pro comprises 1,581 instructions across 23 applications, 5 industries, and 3 operating systems, using high-resolution images from professional domains with expert annotations.
It measures the GUI grounding capabilities of multimodal large language models in professional software scenarios.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of C.
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
Common questions about ScreenSpot Pro.
GPT-6 Astra is currently ranked first with 92.70%.
The current leaders are GPT-6 Astra (92.70%), Claude Opus 4.8 (87.90%), and GPT-5.2 (86.30%).
Qwen3.5-35B-A3B has the lowest matched official input price at $0.25 input / $2.0 output per 1M tokens.
The fastest matched records are Gemini 3 Flash (124.32 tok/s via Google), GPT-5.2 (100.00 tok/s via OpenAI), and Gemini 3 Pro (90.00 tok/s via Google).
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.
It measures the GUI grounding capabilities of multimodal large language models in professional software scenarios.
Yes. Higher values rank better for this benchmark.
26 model results are currently shown.
No. This benchmark is shown for reference but does not contribute to the overall score.
Ranking basisThis screenspot pro AI model leaderboard uses descending score in the benchmark original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
Selection summary
GPT-6 Astra currently leads ScreenSpot Pro with 92.70%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price, and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.