ScreenSpot-Pro
The official leaderboard hasn’t scored a model released after GPT-5 (August 2025). Labs report their own numbers for 29 newer models, such as GPT-6 Astra at 92.7%.
GUI grounding on high-resolution professional-software screenshots. The model must locate the target UI element.
Astra score known only from secondary write-ups of OpenAI launch materials.
- Made by
- ScreenSpot-Pro authors
- Released
- January 2025
- Announced on X
- 925 views, 6 likes
- Kind
- Finding things on screen
- Level
- University
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- ScreenSpot
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last in a launch post
- November 2026
- Last model added
- November 2026
- Scores refreshed
- November 18th, 2026
In the authors’ words
Recent advancements in Multi-modal Large Language Models (MLLMs) have led to significant progress in developing GUI agents for general tasks such as web browsing and mobile phone use. However, their application in professional domains remains under-explored. These specialized workflows introduce unique challenges for GUI perception models, including high-resolution displays, smaller target sizes, and complex environments. In this paper, we introduce ScreenSpot-Pro, a new benchmark designed to rigorously evaluate the grounding capabilities of MLLMs in high-resolution professional settings. The benchmark comprises authentic high-resolution images from a variety of professional domains with expert annotations. It spans 23 applications across five industries and three operating systems. Existing GUI grounding models perform poorly on this dataset, with the best model achieving only 18.9%. Our experiments reveal that strategically reducing the search area enhances accuracy. Based on this insight, we propose ScreenSeekeR, a visual search method that utilizes the GUI knowledge of a strong planner to guide a cascaded search, achieving state-of-the-art performance with 48.1% without any additional training. We hope that our benchmark and findings will advance the development of GUI agents for professional applications. Code, data and leaderboard can be found at https://gui-agent.github.io/grounding-leaderboard.
GPT-6 Astra leads with 92.7%
Best score per model as labs report them in their own cards, drawn striped. Labs pick their own settings, so these compare less cleanly than independent runs.
UI-TARS-1.5 leads with 61.6% across all independent runs
Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Dotted bars are reruns by Epoch AI or another board, and each row names its source.
GPT-6 Astra leads with 92.7%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 92.7 GPT-6 Astra
- 87.9 Claude Opus 4.8
- 87.6 Claude Opus 4.7
- 86.3 GPT-5.2
- 84.5
Qwen 3.8 Max
- 84.1
Muse Spark
- 83.8
Inkling 2 Small
- 79.5 Claude Mythos Preview
- 79.0
Qwen3.7-Plus
- 76.5
Inkling 2 Mini
- 75.4
Muse Glimmer
- 73.1
Seed1.8
- 72.7
Gemini 3 Pro
- 70.4
Qwen3.5-122B-A10B
- 70.3
Qwen3.5-27B
- 69.1
Gemini 3 Flash
- 68.6
Qwen3.5-35B-A3B
- 68.2
Qwen 3.6 Plus
- 66.8
Fara1.5-27B
- 65.6
Qwen3.5 397B-A17B
- 65.2
Qwen3.5-9B
- 61.8
Qwen3-VL-235B-A22B
- 61.6
UI-TARS-1.5
- 60.9
Seed1.5-VL
- 60.3
Qwen3.5-4B
- 57.8
Nemotron 3 Nano Omni 30B-A3B
- 54.5
Qwen3.5-2B
- 52.8 Kimi-VL-A3B-Thinking-2506
- 51.5
STEP3-VL-10B
- 46.5
Qwen3.5-0.8B
- 43.6
Qwen2.5-VL-72B
- 43.5
Nemotron 3.5 Super VL
- 5.50
Nemotron Nano 12B v2 VL
- 1 GPT-6 Astra 92.7%
- 2 Claude Opus 4.8 87.9%
- 3 Claude Opus 4.7 87.6%
- 4 GPT-5.2 86.3%
- 5
Qwen 3.8 Max 84.5%
- 6
Muse Spark 84.1%
- 7
Inkling 2 Small 83.8%
- 8 Claude Mythos Preview max effort 79.5%
- 9
Qwen3.7-Plus 79%
- 10
Inkling 2 Mini 76.5%
- 11
Muse Glimmer 75.4%
- 12
Seed1.8 73.1%
- 13
Gemini 3 Pro 72.7%
- 14
Qwen3.5-122B-A10B 70.4%
- 15
Qwen3.5-27B 70.3%
- 16
Gemini 3 Flash 69.1%
- 17
Qwen3.5-35B-A3B 68.6%
- 18
Qwen 3.6 Plus 68.2%
- 19
Fara1.5-27B 66.8%
- 20
Qwen3.5 397B-A17B 65.6%
- 21
Qwen3.5-9B 65.2%
- 22
Qwen3-VL-235B-A22B 61.8%
- 23
UI-TARS-1.5 61.6%
- 24
Seed1.5-VL 60.9%
- 25
Qwen3.5-4B 60.3%
- 26
Nemotron 3 Nano Omni 30B-A3B 57.8%
- 27
Qwen3.5-2B 54.5%
- 28 Kimi-VL-A3B-Thinking-2506 52.8%
- 29
STEP3-VL-10B 51.5%
- 30
Qwen3.5-0.8B 46.5%
- 31
Qwen2.5-VL-72B 43.6%
- 32
Nemotron 3.5 Super VL 43.5%
- 33
Nemotron Nano 12B v2 VL 5.50%
Source: Labs’ model cards
- 1
UI-TARS-1.5 61.6%
- 2
Seed1.5-VL 60.9%
- 3
Qwen2.5-VL-72B 53.3%
- 4 Kimi-VL-A3B-Thinking-2506 51%
- 5
Qwen2.5-VL-32B 48%
- 6
Qwen2.5-VL-7B 26.8%
- 7 GPT-5 minimal; resized screenshots 18.5%
- 8
Qwen2.5-VL-3B 16.1%
- 9 CogAgent 7.72%
- 10 Qwen2-VL-7B-Instruct 1.58%
- 11 Qwen2-VL-72B-Instruct 1.01%
Source: Official ScreenSpot-Pro leaderboard
- 92.7 GPT-6 Astra
- 87.9 Claude Opus 4.8
- 87.6 Claude Opus 4.7
- 86.3 GPT-5.2
- 84.5
Qwen 3.8 Max
- 84.1
Muse Spark
- 83.8
Inkling 2 Small
- 79.5 Claude Mythos Preview
- 79.0
Qwen3.7-Plus
- 76.5
Inkling 2 Mini
- 75.4
Muse Glimmer
- 73.1
Seed1.8
- 72.7
Gemini 3 Pro
- 70.4
Qwen3.5-122B-A10B
- 70.3
Qwen3.5-27B
- 69.1
Gemini 3 Flash
- 68.6
Qwen3.5-35B-A3B
- 68.2
Qwen 3.6 Plus
- 66.8
Fara1.5-27B
- 65.6
Qwen3.5 397B-A17B
- 65.2
Qwen3.5-9B
- 61.8
Qwen3-VL-235B-A22B
- 61.6
UI-TARS-1.5
- 60.9
Seed1.5-VL
- 60.3
Qwen3.5-4B
- 57.8
Nemotron 3 Nano Omni 30B-A3B
- 54.5
Qwen3.5-2B
- 53.3
Qwen2.5-VL-72B
- 52.8 Kimi-VL-A3B-Thinking-2506
- 51.5
STEP3-VL-10B
- 48.0
Qwen2.5-VL-32B
- 46.5
Qwen3.5-0.8B
- 43.5
Nemotron 3.5 Super VL
- 26.8
Qwen2.5-VL-7B
- 18.5 GPT-5
- 16.1
Qwen2.5-VL-3B
- 7.72 CogAgent
- 5.50
Nemotron Nano 12B v2 VL
- 1.58 Qwen2-VL-7B-Instruct
- 1.01 Qwen2-VL-72B-Instruct
- 1 GPT-6 Astra 92.7%
- 2 Claude Opus 4.8 87.9%
- 3 Claude Opus 4.7 87.6%
- 4 GPT-5.2 86.3%
- 5
Qwen 3.8 Max 84.5%
- 6
Muse Spark 84.1%
- 7
Inkling 2 Small 83.8%
- 8 Claude Mythos Preview max effort 79.5%
- 9
Qwen3.7-Plus 79%
- 10
Inkling 2 Mini 76.5%
- 11
Muse Glimmer 75.4%
- 12
Seed1.8 73.1%
- 13
Gemini 3 Pro 72.7%
- 14
Qwen3.5-122B-A10B 70.4%
- 15
Qwen3.5-27B 70.3%
- 16
Gemini 3 Flash 69.1%
- 17
Qwen3.5-35B-A3B 68.6%
- 18
Qwen 3.6 Plus 68.2%
- 19
Fara1.5-27B 66.8%
- 20
Qwen3.5 397B-A17B 65.6%
- 21
Qwen3.5-9B 65.2%
- 22
Qwen3-VL-235B-A22B 61.8%
- 23
UI-TARS-1.5 61.6%
- 24
Seed1.5-VL 60.9%
- 25
Qwen3.5-4B 60.3%
- 26
Nemotron 3 Nano Omni 30B-A3B 57.8%
- 27
Qwen3.5-2B 54.5%
- 28
Qwen2.5-VL-72B 53.3%
- 29 Kimi-VL-A3B-Thinking-2506 52.8%
- 30
STEP3-VL-10B 51.5%
- 31
Qwen2.5-VL-32B 48%
- 32
Qwen3.5-0.8B 46.5%
- 33
Nemotron 3.5 Super VL 43.5%
- 34
Qwen2.5-VL-7B 26.8%
- 35 GPT-5 minimal; resized screenshots 18.5%
- 36
Qwen2.5-VL-3B 16.1%
- 37 CogAgent 7.72%
- 38
Nemotron Nano 12B v2 VL 5.50%
- 39 Qwen2-VL-7B-Instruct 1.58%
- 40 Qwen2-VL-72B-Instruct 1.01%
Sources: Labs’ model cards and official ScreenSpot-Pro leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 92.7% | Sep 2026 | Lab model card | |
| 2 | Claude Opus 4.8 | 87.9% | May 2026 | Lab model card | |
| 3 | Claude Opus 4.7 | 87.6% | Apr 2026 | Lab model card | |
| 4 | GPT-5.2 | 86.3% | Dec 2025 | Lab model card | |
| 5 | | 84.5% | Aug 2026 | Lab model card | |
| 6 | | 84.1% | Apr 2026 | Lab model card | |
| 7 | | 83.8% | Nov 2026 | Lab model card | |
| 8 | Claude Mythos Preview | 79.5% | max effort | Apr 2026 | Lab model card |
| 9 | | 79% | Jun 2026 | Lab model card | |
| 10 | | 76.5% | Nov 2026 | Lab model card | |
| 11 | | 75.4% | Aug 2026 | Lab model card | |
| 12 | | 73.1% | Dec 2025 | Lab model card | |
| 13 | | 72.7% | Nov 2025 | Lab model card | |
| 14 | | 70.4% | Feb 2026 | Lab model card | |
| 15 | | 70.3% | Feb 2026 | Lab model card | |
| 16 | | 69.1% | Dec 2025 | Lab model card | |
| 17 | | 68.6% | Feb 2026 | Lab model card | |
| 18 | | 68.2% | Apr 2026 | Lab model card | |
| 19 | | 66.8% | Jun 2026 | Lab model card | |
| 20 | | 65.6% | Feb 2026 | Lab model card | |
| 21 | | 65.2% | Mar 2026 | Lab model card | |
| 22 | | 61.8% | Sep 2025 | Lab model card | |
| 23 | | 61.6% | Apr 2025 | Lab model card | |
| 24 | | 60.9% | May 2025 | Lab model card | |
| 25 | | 60.3% | Mar 2026 | Lab model card | |
| 26 | | 57.8% | Apr 2026 | Lab model card | |
| 27 | | 54.5% | Mar 2026 | Lab model card | |
| 28 | Kimi-VL-A3B-Thinking-2506 | 52.8% | Jun 2025 | Lab model card | |
| 29 | | 51.5% | Jan 2026 | Lab model card | |
| 30 | | 46.5% | Mar 2026 | Lab model card | |
| 31 | | 43.6% | Jan 2025 | Lab model card | |
| 32 | | 43.5% | Oct 2026 | Lab model card | |
| 33 | | 5.50% | Nov 2025 | Lab model card |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | 61.6% | Apr 2025 | Official leaderboard | ||
| 2 | 60.9% | May 2025 | Official leaderboard | ||
| 3 | 53.3% | Jan 2025 | Official leaderboard | ||
| 4 | Kimi-VL-A3B-Thinking-2506 | 51% | Jun 2025 | Official leaderboard | |
| 5 | 48% | Mar 2025 | Official leaderboard | ||
| 6 | 26.8% | Jan 2025 | Official leaderboard | ||
| 7 | GPT-5 | 18.5% | minimal; resized screenshots | Aug 2025 | Official leaderboard |
| 8 | 16.1% | Jan 2025 | Official leaderboard | ||
| 9 | CogAgent | 7.72% | Dec 2023 | Official leaderboard | |
| 10 | Qwen2-VL-7B-Instruct | 1.58% | Aug 2024 | Official leaderboard | |
| 11 | Qwen2-VL-72B-Instruct | 1.01% | Aug 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 92.7% | Sep 2026 | Lab model card | |
| 2 | Claude Opus 4.8 | 87.9% | May 2026 | Lab model card | |
| 3 | Claude Opus 4.7 | 87.6% | Apr 2026 | Lab model card | |
| 4 | GPT-5.2 | 86.3% | Dec 2025 | Lab model card | |
| 5 | 84.5% | Aug 2026 | Lab model card | ||
| 6 | 84.1% | Apr 2026 | Lab model card | ||
| 7 | 83.8% | Nov 2026 | Lab model card | ||
| 8 | Claude Mythos Preview | 79.5% | max effort | Apr 2026 | Lab model card |
| 9 | 79% | Jun 2026 | Lab model card | ||
| 10 | 76.5% | Nov 2026 | Lab model card | ||
| 11 | 75.4% | Aug 2026 | Lab model card | ||
| 12 | 73.1% | Dec 2025 | Lab model card | ||
| 13 | 72.7% | Nov 2025 | Lab model card | ||
| 14 | 70.4% | Feb 2026 | Lab model card | ||
| 15 | 70.3% | Feb 2026 | Lab model card | ||
| 16 | 69.1% | Dec 2025 | Lab model card | ||
| 17 | 68.6% | Feb 2026 | Lab model card | ||
| 18 | 68.2% | Apr 2026 | Lab model card | ||
| 19 | 66.8% | Jun 2026 | Lab model card | ||
| 20 | 65.6% | Feb 2026 | Lab model card | ||
| 21 | 65.2% | Mar 2026 | Lab model card | ||
| 22 | 61.8% | Sep 2025 | Lab model card | ||
| 23 | 61.6% | Apr 2025 | Official leaderboard | ||
| 24 | 60.9% | May 2025 | Official leaderboard | ||
| 25 | 60.3% | Mar 2026 | Lab model card | ||
| 26 | 57.8% | Apr 2026 | Lab model card | ||
| 27 | 54.5% | Mar 2026 | Lab model card | ||
| 28 | 53.3% | Jan 2025 | Official leaderboard | ||
| 29 | Kimi-VL-A3B-Thinking-2506 | 52.8% | Jun 2025 | Lab model card | |
| 30 | 51.5% | Jan 2026 | Lab model card | ||
| 31 | 48% | Mar 2025 | Official leaderboard | ||
| 32 | 46.5% | Mar 2026 | Lab model card | ||
| 33 | 43.5% | Oct 2026 | Lab model card | ||
| 34 | 26.8% | Jan 2025 | Official leaderboard | ||
| 35 | GPT-5 | 18.5% | minimal; resized screenshots | Aug 2025 | Official leaderboard |
| 36 | 16.1% | Jan 2025 | Official leaderboard | ||
| 37 | CogAgent | 7.72% | Dec 2023 | Official leaderboard | |
| 38 | 5.50% | Nov 2025 | Lab model card | ||
| 39 | Qwen2-VL-7B-Instruct | 1.58% | Aug 2024 | Official leaderboard | |
| 40 | Qwen2-VL-72B-Instruct | 1.01% | Aug 2024 | Official leaderboard |
The benchmark's own leaderboard has 95 entries, 11 of them models we track. Switch the scores above to rank them together.
Rankings by part
We read 8 parts from the benchmark’s own leaderboard in 2 groups: category and domain. Each ranks models on one part of the test. UI-TARS-1.5 leads 4 of them, and Seed1.5-VL leads the other 4. Arrows show how far a model moved from its overall rank among the models we track.
Icon
11 models we track have a score here.
- 1
UI-TARS-1.5 37.9%
- 2
Seed1.5-VL 32.9%
- 3
Qwen2.5-VL-72B 27.3%
- 4 Kimi-VL-A3B-Thinking-2506 25.3%
- 5
Qwen2.5-VL-32B 24.3%
- 6
Qwen2.5-VL-7B 7.12%
- 7 GPT-5 5.79%
- 8
Qwen2.5-VL-3B 3.81%
- 9 CogAgent 0.83%
- 10 Qwen2-VL-72B-Instruct ↑1 0.33%
Text
11 models we track have a score here.
- 1
Seed1.5-VL ↑1 78.2%
- 2
UI-TARS-1.5 ↓1 76.3%
- 3
Qwen2.5-VL-72B 69.4%
- 4 Kimi-VL-A3B-Thinking-2506 66.8%
- 5
Qwen2.5-VL-32B 62.6%
- 6
Qwen2.5-VL-7B 38.9%
- 7 GPT-5 26.4%
- 8
Qwen2.5-VL-3B 23.6%
- 9 CogAgent 12%
- 10 Qwen2-VL-7B-Instruct 2.46%
Creative
11 models we track have a score here.
- 1
Seed1.5-VL ↑1 59.2%
- 2
UI-TARS-1.5 ↓1 50.4%
- 3
Qwen2.5-VL-72B 44.9%
- 4 Kimi-VL-A3B-Thinking-2506 44.6%
- 5
Qwen2.5-VL-32B 42.2%
- 6
Qwen2.5-VL-7B 24%
- 7 GPT-5 18.5%
- 8
Qwen2.5-VL-3B 16.4%
- 9 CogAgent 5.57%
- 10 Qwen2-VL-7B-Instruct 0.88%
CAD
11 models we track have a score here.
- 1
Seed1.5-VL ↑1 59%
- 2
UI-TARS-1.5 ↓1 58.2%
- 3 Kimi-VL-A3B-Thinking-2506 ↑1 54%
- 4
Qwen2.5-VL-72B ↓1 44.4%
- 5
Qwen2.5-VL-32B 31%
- 6 GPT-5 ↑1 14.2%
- 7
Qwen2.5-VL-7B ↓1 13%
- 8
Qwen2.5-VL-3B 7.28%
- 9 CogAgent 6.13%
- 10 Qwen2-VL-72B-Instruct ↑1 0.77%
Scientific
11 models we track have a score here.
- 1
UI-TARS-1.5 69.3%
- 2
Seed1.5-VL 61.4%
- 3
Qwen2.5-VL-72B 59.1%
- 4
Qwen2.5-VL-32B ↑1 55.5%
- 5 Kimi-VL-A3B-Thinking-2506 ↓1 55.1%
- 6
Qwen2.5-VL-7B 31.1%
- 7 GPT-5 28%
- 8
Qwen2.5-VL-3B 24.8%
- 9 CogAgent 13.4%
- 10 Qwen2-VL-7B-Instruct 3.54%
Office
11 models we track have a score here.
- 1
UI-TARS-1.5 79.6%
- 2
Seed1.5-VL 74.8%
- 3
Qwen2.5-VL-72B 72.6%
- 4 Kimi-VL-A3B-Thinking-2506 67.4%
- 5
Qwen2.5-VL-32B 64.3%
- 6
Qwen2.5-VL-7B 45.2%
- 7
Qwen2.5-VL-3B ↑1 29.6%
- 8 GPT-5 ↓1 23%
- 9 CogAgent 10%
- 10 Qwen2-VL-7B-Instruct 3.04%
OS
11 models we track have a score here.
- 1
Seed1.5-VL ↑1 60.2%
- 2
UI-TARS-1.5 ↓1 51%
- 3
Qwen2.5-VL-32B ↑2 50.5%
- 4
Qwen2.5-VL-72B ↓1 49.5%
- 5 Kimi-VL-A3B-Thinking-2506 ↓1 44.4%
- 6
Qwen2.5-VL-7B 23.5%
- 7 GPT-5 16.3%
- 8
Qwen2.5-VL-3B 6.12%
- 9 CogAgent 3.06%
- 10 Qwen2-VL-72B-Instruct ↑1 0.51%
Development and programming
11 models we track have a score here.
- 1
UI-TARS-1.5 63.9%
- 2
Seed1.5-VL 53.8%
- 3
Qwen2.5-VL-72B 53.5%
- 4
Qwen2.5-VL-32B ↑1 48.8%
- 5 Kimi-VL-A3B-Thinking-2506 ↓1 43.8%
- 6
Qwen2.5-VL-7B 26.1%
- 7 GPT-5 12.4%
- 8
Qwen2.5-VL-3B 12%
- 9 CogAgent 8.03%
- 10 Qwen2-VL-7B-Instruct 1.34%
8 models have held number one in labs’ own numbers since its release in January 2025 3 models have held number one across all independent runs since its release in January 2025 9 models have held number one with claims included since its release in January 2025
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official ScreenSpot-Pro leaderboard and labs’ model cards
OpenAI’s own number for GPT-5.2 passed 85% in December 2025
That was 11 months after its release. By the same age, 20% of the benchmarks from 2025 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official ScreenSpot-Pro leaderboard and labs’ model cards
4 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Nov 20, 2026 | 76.5 | ScreenSpot Pro (original, 87.7 with Python) | |
| Nov 5, 2026 | 83.8 | ScreenSpot Pro (original; 91.5 with Python) | |
| Sep 3, 2026 | GPT-6 Astra | 92.7 | ScreenSpot-Pro (no tools) |
| Aug 10, 2026 | 75.4 | ScreenSpot Pro | |
| Aug 3, 2026 | 84.5 | ScreenSpot Pro | |
| Jun 1, 2026 | 79 | ScreenSpot Pro (enable_thinking=False) | |
| May 28, 2026 | Claude Opus 4.8 | 87.9 | ScreenSpot-Pro (with tools) |
| Apr 16, 2026 | Claude Opus 4.7 | 87.6 | ScreenSpot-Pro (with tools) |
| Apr 8, 2026 | 84.1 | ScreenSpot Pro (Screenshot Localization - With Python) | |
| Apr 2, 2026 | 68.2 | ScreenSpot Pro | |
| Mar 2, 2026 | 46.5 | ScreenSpot Pro --/46.5 | |
| Mar 2, 2026 | 54.5 | ScreenSpot Pro --/54.5 | |
| Mar 2, 2026 | 60.3 | ScreenSpot Pro | |
| Mar 2, 2026 | 65.2 | ScreenSpot Pro | |
| Feb 24, 2026 | 68.6 | ScreenSpot Pro | |
| Feb 24, 2026 | 70.3 | ScreenSpot Pro | |
| Feb 24, 2026 | 70.4 | ScreenSpot Pro | |
| Feb 16, 2026 | 65.6 | ScreenSpot Pro | |
| Jan 13, 2026 | 51.5 | ScreenSpot-Pro | |
| Dec 18, 2025 | 73.1 | ScreenSpot-Pro 64.3 / 73.1 with crop-box tool | |
| Dec 17, 2025 | 69.1 | ScreenSpot-Pro (No tools unless specified) | |
| Dec 11, 2025 | GPT-5.2 | 86.3 | ScreenSpot-Pro GUI screenshot understanding (w/ Python) |
| Nov 18, 2025 | 72.7 | ScreenSpot-Pro | |
| Sep 23, 2025 | 61.8 | ScreenSpot Pro | |
| Jun 21, 2025 | Kimi-VL-A3B-Thinking-2506 | 52.8 | ScreenSpot-Pro (Acc) |
| May 12, 2025 | 60.9 | ScreenSpot-Pro | |
| Apr 16, 2025 | 61.6 | ScreenSpotPro | |
| Jan 28, 2025 | 43.6 | ScreenSpot Pro |
15 system cards report it, from 9 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 20, 2026 | | 76.5 | ScreenSpot Pro (original, 87.7 with Python) | Capabilities |
| Oct 19, 2026 | | 75.4 | Screenspot Pro, high (Tables 2 and 5) | Capabilities |
| Oct 15, 2026 | | 43.5 | ScreenSpot Pro, BF16 acc. (Table 23) | Capabilities |
| Jun 18, 2026 | | 66.8 | ScreenSpot-Pro with two-step zoom, 58.6 without zoom (Table 6) | Capabilities |
| May 28, 2026 | Claude Opus 4.8 system card | 82.3 | ScreenSpot-Pro, no tools, max effort (87.9% with Python tools) (Table 8.1.A, 8.12.5) | Capabilities |
| Apr 27, 2026 | | 57.8 | ScreenSpot-Pro, reasoning on, 59.3 reasoning off (Table 7) | Capabilities |
| Apr 16, 2026 | Claude Opus 4.7 system card | 79.5 | ScreenSpot-Pro, no tools. 87.6 with Python tools (Table 8.1.A, 8.9.3) | Capabilities |
| Apr 7, 2026 | Claude Mythos Preview system card | 79.5 | ScreenSpot-Pro, adaptive thinking, max effort, no tools, average of five runs (6.11.2); 92.8 with Python tools | Capabilities |
| Mar 21, 2026 | | 64.3 | ScreenSpot-Pro 64.3, 73.1 with the crop-box tool (Table 2) | Capabilities |
| Jan 14, 2026 | | 51.5 | ScreenSpot-Pro, SeRe (Table 1) | Capabilities |
| Nov 26, 2025 | | 61.8 | ScreenSpot Pro (Table 2) | Capabilities |
| Nov 6, 2025 | | 5.5 | ScreenSpot-Pro, reasoning on (Table 3). Reasoning off 4.8 | Capabilities |
| Jun 23, 2025 | Kimi-VL-A3B-Thinking-2506 technical report | 52.8 | ScreenSpot-Pro (Acc) (Table 5) | Capabilities |
| May 11, 2025 | | 60.9 | ScreenSpot-Pro, thinking (Table 8) | Capabilities |
| Feb 19, 2025 | | 43.6 | ScreenSpot Pro (Table 9) | Capabilities |
The lab's number comes out ahead in 1 of 2 pairs that differ
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 2 of them match, which usually means one side copied the other. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| | 43.6% ScreenSpot Pro | 53.3% Leaderboard | −9.7 pts |
| Kimi-VL-A3B-Thinking-2506 | 52.8% ScreenSpot-Pro (Acc) | 51% Leaderboard | +1.8 pts |
| | 61.6% ScreenSpotPro | 61.6% Leaderboard | Same |
| | 60.9% ScreenSpot-Pro | 60.9% Leaderboard | Same |
More in Computer and browser use
- BrowseComp In use Hard-to-find facts that need persistent web browsing. Short verifiable answers graded for exact correctness. Claude Fable 5 93.3%
- AndroidWorld In use Agents complete 116 tasks across 20 real Android apps in an emulator. Each task checks the device's system state to judge success. GPT-6 Astra 100%
- OSWorld 2.1 In use Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. Claude Opus 5 44.3%
- WebArena-Verified In use Agents complete 812 audited WebArena tasks on self-hosted websites. Deterministic evaluators check the agent's answer and captured network traces.
Muse Spark 1.1 69%
- OSWorld 2.1 (offline subset) In use Agents complete the 82 OSWorld 2.1 tasks on the official offline list, with no internet access. Scored by checkpoint partial credit and strict pass rate. Claude Fable 5.2 89.3%
- MobileWorld In use Agents complete 201 tasks across 20 mobile apps, including tasks that need user clarification or MCP tools. Backend checks verify success. GPT-5 55.3%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (95 entries), launch posts (28 scores) and system cards (15 scores).
- Pass Models scored 40 models we track, plus 84 other entries on its leaderboard.
- Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
- Note Frontier models 1 of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-6 Astra (#4).
- Note Headroom Saturated. The best is 92.7%, past 85% since December 2025.
Leaderboard import 3 pass
- Pass Importer We read the board on gui-agent.github.io: 12 rows for models we track and 84 other entries at the last read.
- Pass Last read November 18th, 2026.
- Pass Read worked The last read worked.
Consistency 2 pass
- Pass Lab claims All 4 pairs agree within 10 points.
- Pass Known issues No published audit found problems.