GUI-CEval
Models answer questions about Chinese app screens, locate controls, and choose actions. Agents also complete tasks on physical devices, where the grader checks interface elements.
Xiaomi tested 20 models and agent systems across 201 apps. Its conference supplement includes annotations and evaluation code but only some screenshots.
- Made by
- Xiaomi HyperAI Team
- Released
- March 2026
- Kind
- Phones
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- Chinese
- Interaction
- Agentic
- Environment
- Phone apps, A physical robot or device
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Computer use, Agents, Robotics
- Last model added
- May 2025
- Scores refreshed
- November 20th, 2026
Qwen2.5-VL-72B leads with 61.4% among models released by May 2025 Qwen2.5-VL-72B leads with 61.4% among models released by May 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in May 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 61.4
Qwen2.5-VL-72B
- 56.2 UI-TARS-72B-SFT
- 55.5
Qwen2.5-VL-32B
- 53.7
Qwen2.5-VL-7B
- 50.0 UI-TARS-7B-SFT
- 49.7 MiMo-VL-7B-RL
- 48.7
Qwen2.5-VL-3B
- 46.6 MIMO-VL-SFT
- 43.9
UI-TARS-72B-DPO
- 39.5 UI-TARS-2B-SFT
- 30.0 UI-TARS-7B-DPO
- 30.0 TongUI-7B
- 28.8 OS-Atlas-base-7B
- 27.7 GPT-4o (snapshot unspecified)
- 25.3 TongUI-3B
- 20.9 GPT-4o-mini (snapshot unspecified)
- 1
Qwen2.5-VL-72B 61.4%
- 2 UI-TARS-72B-SFT 56.2%
- 3
Qwen2.5-VL-32B 55.5%
- 4
Qwen2.5-VL-7B 53.7%
- 5 UI-TARS-7B-SFT 50%
- 6 MiMo-VL-7B-RL 49.7%
- 7
Qwen2.5-VL-3B 48.7%
- 8 MIMO-VL-SFT 46.6%
- 9
UI-TARS-72B-DPO 43.9%
- 10 UI-TARS-2B-SFT 39.5%
- 11 UI-TARS-7B-DPO 30%
- 12 TongUI-7B 30%
- 13 OS-Atlas-base-7B 28.8%
- 14 GPT-4o (snapshot unspecified) 27.7%
- 15 TongUI-3B 25.3%
- 16 GPT-4o-mini (snapshot unspecified) 20.9%
Source: The GUI-CEval paper
- 1
Qwen2.5-VL-72B 61.4%
- 2
Qwen2.5-VL-32B 55.5%
- 3
Qwen2.5-VL-7B 53.7%
- 4 MiMo-VL-7B-RL 49.7%
- 5
Qwen2.5-VL-3B 48.7%
- 6
UI-TARS-72B-DPO 43.9%
Source: The GUI-CEval paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 61.4% | Jan 2025 | Official leaderboard | |
| 2 | UI-TARS-72B-SFT | 56.2% | Official leaderboard | ||
| 3 | | 55.5% | Mar 2025 | Official leaderboard | |
| 4 | | 53.7% | Jan 2025 | Official leaderboard | |
| 5 | UI-TARS-7B-SFT | 50% | Official leaderboard | ||
| 6 | MiMo-VL-7B-RL | 49.7% | May 2025 | Official leaderboard | |
| 7 | | 48.7% | Jan 2025 | Official leaderboard | |
| 8 | MIMO-VL-SFT | 46.6% | Official leaderboard | ||
| 9 | | 43.9% | Jan 2025 | Official leaderboard | |
| 10 | UI-TARS-2B-SFT | 39.5% | Official leaderboard | ||
| 11 | UI-TARS-7B-DPO | 30% | Official leaderboard | ||
| 12 | TongUI-7B | 30% | Official leaderboard | ||
| 13 | OS-Atlas-base-7B | 28.8% | Official leaderboard | ||
| 14 | GPT-4o (snapshot unspecified) | 27.7% | Official leaderboard | ||
| 15 | TongUI-3B | 25.3% | Official leaderboard | ||
| 16 | GPT-4o-mini (snapshot unspecified) | 20.9% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 61.4% | Jan 2025 | Official leaderboard | |
| 2 | | 55.5% | Mar 2025 | Official leaderboard | |
| 3 | | 53.7% | Jan 2025 | Official leaderboard | |
| 4 | MiMo-VL-7B-RL | 49.7% | May 2025 | Official leaderboard | |
| 5 | | 48.7% | Jan 2025 | Official leaderboard | |
| 6 | | 43.9% | Jan 2025 | Official leaderboard |
Rankings by part
We read 7 parts from the benchmark’s paper, each ranking models on one part of the test, such as Perception, Planning and Reflection. Qwen2.5-VL-72B leads 3 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Perception
6 models we track have a score here.
- 1
Qwen2.5-VL-72B 82.3%
- 2
Qwen2.5-VL-7B ↑1 75.3%
- 3
Qwen2.5-VL-32B ↓1 75.3%
- 4 MiMo-VL-7B-RL 73.6%
- 5
Qwen2.5-VL-3B 65.6%
- 6
UI-TARS-72B-DPO 42.5%
Planning
6 models we track have a score here.
- 1
Qwen2.5-VL-72B 66.7%
- 2
Qwen2.5-VL-32B 63.6%
- 3
Qwen2.5-VL-7B 54.8%
- 4 MiMo-VL-7B-RL 51.7%
- 5
Qwen2.5-VL-3B 43%
- 6
UI-TARS-72B-DPO 27.6%
Reflection
6 models we track have a score here.
- 1
Qwen2.5-VL-72B 21%
- 2
Qwen2.5-VL-32B 19.2%
- 3 MiMo-VL-7B-RL ↑1 15.6%
- 4
Qwen2.5-VL-7B ↓1 14.3%
- 5
Qwen2.5-VL-3B 12.7%
- 6
UI-TARS-72B-DPO 6.66%
Evaluation
6 models we track have a score here.
- 1
Qwen2.5-VL-32B ↑1 49.2%
- 2 MiMo-VL-7B-RL ↑2 46.8%
- 3
Qwen2.5-VL-72B ↓2 40.1%
- 4
UI-TARS-72B-DPO ↑2 36.6%
- 5
Qwen2.5-VL-7B ↓2 34%
- 6
Qwen2.5-VL-3B ↓1 23%
Grounding
6 models we track have a score here.
- 1 MiMo-VL-7B-RL ↑3 90%
- 2
Qwen2.5-VL-7B ↑1 89.4%
- 3
Qwen2.5-VL-32B ↓1 88.7%
- 4
Qwen2.5-VL-72B ↓3 88.1%
- 5
Qwen2.5-VL-3B 83.2%
- 6
UI-TARS-72B-DPO 75.7%
Offline Agent
6 models we track have a score here.
- 1
UI-TARS-72B-DPO ↑5 74.4%
- 2
Qwen2.5-VL-72B ↓1 70.3%
- 3
Qwen2.5-VL-32B ↓1 70%
- 4
Qwen2.5-VL-3B ↑1 64.6%
- 5 MiMo-VL-7B-RL ↓1 60.8%
- 6
Qwen2.5-VL-7B ↓3 54.2%
Online Agent
6 models we track have a score here.
- 1
Qwen2.5-VL-32B ↑1 31.9%
- 2
UI-TARS-72B-DPO ↑4 29.2%
- 3
Qwen2.5-VL-3B ↑2 27.7%
- 4
Qwen2.5-VL-72B ↓3 26.9%
- 5 MiMo-VL-7B-RL ↓1 17.2%
- 6
Qwen2.5-VL-7B ↓3 9.17%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The GUI-CEval paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official GUI-CEval leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
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%
- ScreenSpot-Pro In use GUI grounding on high-resolution professional-software screenshots. The model must locate the target UI element. GPT-6 Astra 92.7%
- 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%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 18 months ago, in May 2025, for MiMo-VL-7B-RL. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (16 entries).
- Pass Models scored 6 models we track, plus 10 other entries on its leaderboard.
- Warning Newest score The newest came 18 months ago, in May 2025, for MiMo-VL-7B-RL.
- Note Frontier models None of its models are on the capabilities index.
- Pass Headroom The best is 61.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on openaccess.thecvf.com: 6 rows for models we track and 10 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.