benchgap
Vision & documents

ScreenSpot Pro leaderboard

As of 2026-11-21, the highest measured score on ScreenSpot Pro is 92.7% by GPT-6 Astra. 139 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Opus 5.595.3%estimated ± 10.1 pp, low confidence
2Claude Fable 5.194.6%estimated ± 10.1 pp, low confidence
3GPT-6 Astra92.7%measured
4Gemini 3 Flash92.2%estimated ± 10.1 pp, low confidence
5Gemini 3.8 Flash90.2%estimated ± 10.1 pp, low confidence
6Claude Fable 589.2%estimated ± 10.1 pp, low confidence
7GPT-6.1 Sol89.1%estimated ± 10.1 pp, low confidence
8Claude Sonnet 5.588.8%estimated ± 10.1 pp, low confidence
9Gemini 3.6 Flash88.7%estimated ± 10.1 pp, low confidence
10GPT-6 Sol88.6%estimated ± 10.1 pp, low confidence
11Claude Mythos 588.5%estimated ± 6.8 pp, medium confidence
12Glimpse87.9%estimated ± 6.8 pp, medium confidence
13Claude Opus 4.887.9%measured
14Claude Opus 4.786.1%estimated ± 6.8 pp, medium confidence
15GPT-5.485.4%measured
16LongCat-2.585.3%estimated ± 6.8 pp, medium confidence
17DeepSeek V4.1 Flash 101785.3%estimated ± 10.1 pp, low confidence
18Qwen3.8-Flash-Next85.1%estimated ± 4.8 pp, low confidence
19GPT-6 Luna85.1%estimated ± 10.1 pp, low confidence
20Muse Spark Open85.0%estimated ± 3.6 pp, low confidence
21Kimi K3.1 Pro85.0%estimated ± 3.6 pp, medium confidence
22Qwen3.8 Max84.5%measured
23GLM-5.3-Flash84.5%estimated ± 6.8 pp, medium confidence
24Gemini 3.1 Pro84.4%measured
25Muse Spark84.1%measured
26Inkling 2 Small83.8%measured
27Gemini 3.7 Flash83.7%estimated ± 6.8 pp, medium confidence
28Kimi K383.5%estimated ± 3.6 pp, medium confidence
29Muse Spark 1.183.4%estimated ± 6.8 pp, medium confidence
30Claude Sonnet 583.2%estimated ± 6.8 pp, medium confidence
31Claude Opus 4.683.1%measured
32GPT-6.2 Astra82.9%estimated ± 10.8 pp, low confidence
33Qwen3.8-Omni-Flash82.6%estimated ± 4.8 pp, medium confidence
34Gemini 3.5 Flash-Lite81.4%estimated ± 10.1 pp, low confidence
35Sakana Fugu-Ultra81.2%estimated ± 6.8 pp, medium confidence
36dots3-note Preview80.5%estimated ± 3.6 pp, medium confidence
37Claude Opus 580.2%estimated ± 10.8 pp, low confidence
38GPT-6.1 Terra79.7%estimated ± 9.6 pp, low confidence
39Sakana Fugu79.3%estimated ± 6.8 pp, medium confidence
40Qwen3.8 Max Preview79.1%estimated ± 10.8 pp, low confidence
41Qwen3.7 Plus79.0%measured
42Seed 2.1 Pro78.8%estimated ± 3.6 pp, medium confidence
43Claude Haiku 5.578.6%estimated ± 10.1 pp, low confidence
44GPT-5.6 Sol78.3%estimated ± 9.6 pp, low confidence
45Gemini 3.5 Flash78.0%estimated ± 6.8 pp, medium confidence
46Grok 4.577.7%estimated ± 10.8 pp, low confidence
47GPT-5.577.2%estimated ± 9.6 pp, low confidence
48Apodex 1.177.0%estimated ± 10.8 pp, low confidence
49Apodex 1.1 Mini77.0%estimated ± 10.8 pp, low confidence
50Qwen3.8-27B77.0%estimated ± 4.8 pp, medium confidence
51GPT-5.6 Terra77.0%estimated ± 9.6 pp, low confidence
52Ling 3.0 Flash VL76.9%estimated ± 10.8 pp, low confidence
53GPT-5.3 Codex76.6%estimated ± 10.8 pp, low confidence
54Inkling 2 Mini76.5%measured
55Hy475.7%estimated ± 9.6 pp, low confidence
56GPT-5.6 Luna75.6%estimated ± 9.6 pp, low confidence
57MiniMax M3.1 Flash75.5%estimated ± 6.8 pp, medium confidence
58Grok 4.375.4%estimated ± 9.6 pp, low confidence
59Muse Glimmer 30B75.4%measured
60Pareto 26.975.4%estimated ± 9.6 pp, low confidence
61GPT-5.2-Codex75.3%estimated ± 10.8 pp, low confidence
62Qwen3.6-35B-A3B75.1%estimated ± 4.8 pp, medium confidence
63Inkling74.8%estimated ± 6.8 pp, medium confidence
64GPT-5.174.8%estimated ± 10.8 pp, low confidence
65Claude Opus 4.6 (Adaptive)74.7%estimated ± 10.8 pp, low confidence
66Gemma 4 31B74.7%estimated ± 9.6 pp, low confidence
67GPT-5.4 mini74.5%estimated ± 9.6 pp, low confidence
68Gemini 2.5 Pro74.4%estimated ± 10.8 pp, low confidence
69Claude Haiku 4.5 Thinking74.3%estimated ± 10.1 pp, low confidence
70Step 5 Preview74.1%estimated ± 9.6 pp, low confidence
71Hy4 flash74.0%estimated ± 9.6 pp, low confidence
72GPT-574.0%estimated ± 10.8 pp, low confidence
73Claude Opus 4.5 Thinking73.8%estimated ± 10.8 pp, low confidence
74Inkling-Small73.7%estimated ± 6.8 pp, medium confidence
75MiMo-V2.6-Flash73.3%estimated ± 10.8 pp, low confidence
76MiMo-V2.573.2%estimated ± 6.8 pp, medium confidence
77GLM-5V-Turbo73.1%estimated ± 10.8 pp, low confidence
78Nemotron 3.5 Super VL 120B A12B72.9%estimated ± 6.8 pp, medium confidence
79GPT-5.1-Codex72.9%estimated ± 10.8 pp, low confidence
80GPT-5.1-Codex-Max72.9%estimated ± 10.8 pp, low confidence
81Gemma 4 26B A4B72.8%estimated ± 9.6 pp, low confidence
82Gemini 3 Pro72.7%measured
83Claude Haiku 4.572.6%estimated ± 10.1 pp, low confidence
84Kimi K2.672.2%estimated ± 6.8 pp, medium confidence
85Qwen3.6-27B71.3%estimated ± 4.8 pp, medium confidence
86Nemotron 3.5 Super VL 120B A12B NVFP471.3%estimated ± 6.8 pp, medium confidence
87o371.3%estimated ± 10.8 pp, low confidence
88MiMo-V2-Omni71.2%estimated ± 10.8 pp, low confidence
89Interfaze Beta71.1%estimated ± 9.6 pp, low confidence
90o4-mini70.7%estimated ± 10.8 pp, low confidence
91Qwen3.5-27B70.7%estimated ± 6.8 pp, medium confidence
92Holo2-235B-A22B70.6%measured
93Grok 470.5%estimated ± 10.8 pp, low confidence
94GPT-5.268.8%estimated ± 4.8 pp, medium confidence
95Gemini 2.5 Flash68.2%estimated ± 10.8 pp, low confidence
96Qwen3.6 Plus68.2%measured
97Qwen3.5-4B68.2%estimated ± 10.8 pp, low confidence
98Mistral Medium 3.5 128B67.8%estimated ± 10.8 pp, low confidence
99GPT-5.4 nano67.7%estimated ± 9.6 pp, low confidence
100Qwen3.5-35B-A3B67.1%estimated ± 6.8 pp, medium confidence
101Claude Sonnet 4.666.9%estimated ± 6.8 pp, medium confidence
102Grok 4.1 Fast (Reasoning)66.7%estimated ± 10.8 pp, low confidence
103Qwen3.5-122B-A10B66.6%estimated ± 6.8 pp, medium confidence
104Holo2-30B-A3B66.1%measured
105Claude 4 Sonnet66.1%estimated ± 10.8 pp, low confidence
106Step 3.7 Flash65.9%estimated ± 7.5 pp, low confidence
107Llama 4 Maverick65.9%estimated ± 10.8 pp, low confidence
108Grok 4 Fast (Reasoning)65.7%estimated ± 10.8 pp, low confidence
109Qwen3.5 397B65.6%measured
110GPT-4.165.2%estimated ± 10.8 pp, low confidence
111Qwen3-Omni-30B-A3B-Thinking64.5%estimated ± 10.8 pp, low confidence
112Amazon Nova 2 Lite63.6%estimated ± 10.1 pp, low confidence
113GPT-4.1 mini63.4%estimated ± 10.8 pp, low confidence
114Qwen3-VL-235B-A22B Instruct62.0%measured
115Mistral Small 462.0%estimated ± 10.8 pp, low confidence
116Kimi K2.561.7%estimated ± 4.8 pp, medium confidence
117Mistral Large 361.2%estimated ± 10.8 pp, low confidence
118Qwen3-Omni-30B-A3B-Instruct61.0%estimated ± 10.8 pp, low confidence
119Gemini 1.5 Pro60.6%estimated ± 10.8 pp, low confidence
120Mistral Medium 359.1%estimated ± 10.8 pp, low confidence
121Llama 4 Scout59.0%estimated ± 10.8 pp, low confidence
122Ternary Bonsai 2 27B58.9%estimated ± 4.8 pp, medium confidence
123Holo2-8B58.9%measured
124Gemini 3.1 Flash-Lite58.6%estimated ± 6.8 pp, medium confidence
125Gemma 4 E4B57.9%estimated ± 9.6 pp, low confidence
126Nemotron 3 Nano Omni 30B A3B57.8%measured
127Holo2-4B57.2%measured
128LFM3-VL-2B55.5%measured
129Qwen2.5-VL-32B55.5%estimated ± 9.6 pp, low confidence
130Grok 4.1 Fast55.4%estimated ± 10.8 pp, low confidence
131Claude 4 Opus Thinking55.3%estimated ± 7.9 pp, low confidence
132GLM-4.6V55.3%estimated ± 7.9 pp, low confidence
133Claude 3.7 Sonnet Thinking55.3%estimated ± 7.9 pp, low confidence
134Claude 4 Opus55.3%estimated ± 7.9 pp, low confidence
135Gemini 2.5 Flash-Lite55.3%estimated ± 7.9 pp, low confidence
136Claude 3.7 Sonnet55.3%estimated ± 7.9 pp, low confidence
137GLM-4.6V-Flash55.3%estimated ± 7.9 pp, low confidence
138Gemini 2.0 Flash55.3%estimated ± 7.9 pp, low confidence
139MiniMax-VL-0155.3%estimated ± 7.9 pp, low confidence
140Grok 255.3%estimated ± 7.9 pp, low confidence
141Gemini 2.0 Flash-Lite55.3%estimated ± 7.9 pp, low confidence
142Mistral Small 3.155.3%estimated ± 7.9 pp, low confidence
143Grok 2 mini55.3%estimated ± 7.9 pp, low confidence
144Gemini 1.5 Flash55.3%estimated ± 7.9 pp, low confidence
145LFM2.5-VL-450M55.3%estimated ± 7.9 pp, low confidence
146ZAYA1-VL-8B55.3%estimated ± 7.9 pp, low confidence
147Gemma 3 27B55.0%estimated ± 10.8 pp, low confidence
148Gemma 4 12B53.6%estimated ± 9.3 pp, medium confidence
149Claude 3.5 Haiku53.0%estimated ± 10.8 pp, low confidence
150MiniMax M352.5%estimated ± 8.4 pp, low confidence
151Nova Pro51.9%estimated ± 10.8 pp, low confidence
152Gemma 4 E2B51.0%estimated ± 9.6 pp, low confidence
153GPT-4o mini49.4%estimated ± 10.8 pp, low confidence
154GPT-4.1 nano48.2%estimated ± 10.8 pp, low confidence
155Claude Opus 4.545.7%measured
156Claude 3 Haiku39.2%estimated ± 10.8 pp, low confidence
157LFM2.5-VL-3B37.0%estimated ± 4.8 pp, low confidence
158LFM2.5-VL-1.6B-Extract34.7%estimated ± 10.8 pp, low confidence
159Seed 2.1 Turbo34.0%estimated ± 3.6 pp, low confidence
160Grok 4.2031.1%estimated ± 6.8 pp, low confidence
161Command A+15.8%estimated ± 6.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General