Benchmark Almanac

GUI-CEval

Qwen2.5-VL-72B 61.4% leads since March 2026
Not in model cards

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.

Released
March 2026
Kind
Phones
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Text, Screens → Text, Actions
Interaction
Agentic
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Last model added
May 2025
Scores refreshed
November 20th, 2026
16
Entries in its paper, 6 models we track

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:

  1. 1 Qwen2.5-VL-72B 61.4%
  2. 2 UI-TARS-72B-SFT 56.2%
  3. 3 Qwen2.5-VL-32B 55.5%
  4. 4 Qwen2.5-VL-7B 53.7%
  5. 5 UI-TARS-7B-SFT 50%
  6. 6 MiMo-VL-7B-RL 49.7%
  7. 7 Qwen2.5-VL-3B 48.7%
  8. 8 MIMO-VL-SFT 46.6%
  9. 9 UI-TARS-72B-DPO 43.9%
  10. 10 UI-TARS-2B-SFT 39.5%
  11. 11 UI-TARS-7B-DPO 30%
  12. 12 TongUI-7B 30%
  13. 13 OS-Atlas-base-7B 28.8%
  14. 14 GPT-4o (snapshot unspecified) 27.7%
  15. 15 TongUI-3B 25.3%
  16. 16 GPT-4o-mini (snapshot unspecified) 20.9%

Source: The GUI-CEval paper

#EntryScoreSettingDateSource
1 Qwen2.5-VL-72B 61.4% Jan 2025 Official leaderboard
2 UI-TARS-72B-SFT 56.2% Official leaderboard
3 Qwen2.5-VL-32B 55.5% Mar 2025 Official leaderboard
4 Qwen2.5-VL-7B 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 Qwen2.5-VL-3B 48.7% Jan 2025 Official leaderboard
8 MIMO-VL-SFT 46.6% Official leaderboard
9 UI-TARS-72B-DPO 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
#ModelScoreSettingDateSource
1 Qwen2.5-VL-72B 61.4% Jan 2025 Official leaderboard
2 Qwen2.5-VL-32B 55.5% Mar 2025 Official leaderboard
3 Qwen2.5-VL-7B 53.7% Jan 2025 Official leaderboard
4 MiMo-VL-7B-RL 49.7% May 2025 Official leaderboard
5 Qwen2.5-VL-3B 48.7% Jan 2025 Official leaderboard
6 UI-TARS-72B-DPO 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. 1 Qwen2.5-VL-72B 82.3%
  2. 2 Qwen2.5-VL-7B ↑1 75.3%
  3. 3 Qwen2.5-VL-32B ↓1 75.3%
  4. 4 MiMo-VL-7B-RL 73.6%
  5. 5 Qwen2.5-VL-3B 65.6%
  6. 6 UI-TARS-72B-DPO 42.5%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Qwen
Benchmark paper Not tested since Mar 2026

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.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Qwen2.5-VL-72B UI-TARS-72B-DPO 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Qwen2.5-VL-72B UI-TARS-72B-DPO

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

All 119 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.

How we check