UniGUI-Bench
Agents interpret and operate graphical interfaces. The benchmark tests 11 capabilities and includes separate system evaluations.
- Made by
- AGI-Eval (Meituan)
- Released
- May 2026
- Kind
- Finding things on screen
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Interaction
- Agentic
- Environment
- Desktop apps, Phone apps, Browser
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Computer use, Agents
- Strongest model tested
-
Gemini 3 Flash #60 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Seed-1.8-VL tops the leaderboard with 84.2% Kimi K2.5 leads with 80.9% among models released by April 2026
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. The best model we track is Kimi K2.5, at 80.9%. This view shows only models we track, each at its best.
The newest model with a score here came out in April 2026, 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
- 1 Seed-1.8-VL 84.2%
- 2 Kimi K2.5 80.9%
- 3
Qwen3.5-35B-A3B 78.3%
- 4 GLM-5V-Turbo 78%
- 5 Qwen3-VL-32B 76%
- 6
Gemini 3 Flash 74.9%
- 7
Seed1.5-VL 71.7%
- 8 GLM-4.6V-Flash 66.3%
- 9
Gemini 2.5 Flash Sep 2025 65.2%
- 10 Qwen3-VL-8B 61.9%
- 11
InternVL3-38B 49.2%
- 12
InternVL3.5-8B 43.2%
- 13
InternVL3.5-30B-A3B 42.5%
- 14
InternVL3-8B 35.6%
Source: The UniGUI-Bench paper
- 1 Kimi K2.5 80.9%
- 2
Qwen3.5-35B-A3B 78.3%
- 3 GLM-5V-Turbo 78%
- 4
Gemini 3 Flash 74.9%
- 5
Seed1.5-VL 71.7%
- 6 GLM-4.6V-Flash 66.3%
- 7
Gemini 2.5 Flash Sep 2025 65.2%
- 8
InternVL3-38B 49.2%
- 9
InternVL3.5-8B 43.2%
- 10
InternVL3.5-30B-A3B 42.5%
- 11
InternVL3-8B 35.6%
Source: The UniGUI-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Seed-1.8-VL | 84.2% | Official leaderboard | ||
| 2 | Kimi K2.5 | 80.9% | Jan 2026 | Official leaderboard | |
| 3 | | 78.3% | Feb 2026 | Official leaderboard | |
| 4 | GLM-5V-Turbo | 78% | Apr 2026 | Official leaderboard | |
| 5 | Qwen3-VL-32B | 76% | Official leaderboard | ||
| 6 | | 74.9% | Dec 2025 | Official leaderboard | |
| 7 | | 71.7% | May 2025 | Official leaderboard | |
| 8 | GLM-4.6V-Flash | 66.3% | Dec 2025 | Official leaderboard | |
| 9 | | 65.2% | Sep 2025 | Official leaderboard | |
| 10 | Qwen3-VL-8B | 61.9% | Official leaderboard | ||
| 11 | | 49.2% | Apr 2025 | Official leaderboard | |
| 12 | | 43.2% | Aug 2025 | Official leaderboard | |
| 13 | | 42.5% | Aug 2025 | Official leaderboard | |
| 14 | | 35.6% | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Kimi K2.5 | 80.9% | Jan 2026 | Official leaderboard | |
| 2 | | 78.3% | Feb 2026 | Official leaderboard | |
| 3 | GLM-5V-Turbo | 78% | Apr 2026 | Official leaderboard | |
| 4 | | 74.9% | Dec 2025 | Official leaderboard | |
| 5 | | 71.7% | May 2025 | Official leaderboard | |
| 6 | GLM-4.6V-Flash | 66.3% | Dec 2025 | Official leaderboard | |
| 7 | | 65.2% | Sep 2025 | Official leaderboard | |
| 8 | | 49.2% | Apr 2025 | Official leaderboard | |
| 9 | | 43.2% | Aug 2025 | Official leaderboard | |
| 10 | | 42.5% | Aug 2025 | Official leaderboard | |
| 11 | | 35.6% | Apr 2025 | Official leaderboard |
Rankings by part
We read 11 parts from the benchmark’s paper, each ranking models on one part of the test, such as Element Grounding, Action Understanding and Page Understanding. Kimi K2.5 leads 5 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.
Element Grounding
11 models we track have a score here.
- 1 GLM-5V-Turbo ↑2 94%
- 2
Gemini 3 Flash ↑2 91%
- 3 GLM-4.6V-Flash ↑3 90%
- 4
Qwen3.5-35B-A3B ↓2 88.5%
- 5 Kimi K2.5 ↓4 85%
- 6
Seed1.5-VL ↓1 81.5%
- 7
InternVL3-38B ↑1 62%
- 8
InternVL3.5-8B ↑1 44%
- 9
Gemini 2.5 Flash Sep 2025 ↓2 44%
- 10
InternVL3.5-30B-A3B 34%
Action Understanding
11 models we track have a score here.
- 1 Kimi K2.5 90%
- 2
Gemini 3 Flash ↑2 89.5%
- 3 GLM-5V-Turbo 87.5%
- 4
Qwen3.5-35B-A3B ↓2 85.5%
- 5
Seed1.5-VL 84%
- 6 GLM-4.6V-Flash 76.5%
- 7
Gemini 2.5 Flash Sep 2025 68.5%
- 8
InternVL3-38B 20.5%
- 9
InternVL3.5-8B 20.5%
- 10
InternVL3.5-30B-A3B 17%
Page Understanding
11 models we track have a score here.
- 1
Qwen3.5-35B-A3B ↑1 99.5%
- 2 GLM-5V-Turbo ↑1 98.5%
- 3
Gemini 2.5 Flash Sep 2025 ↑4 97%
- 4 Kimi K2.5 ↓3 97%
- 5
Gemini 3 Flash ↓1 96.5%
- 6
Seed1.5-VL ↓1 95%
- 7 GLM-4.6V-Flash ↓1 93%
- 8
InternVL3-38B 85.5%
- 9
InternVL3.5-30B-A3B ↑1 79%
- 10
InternVL3.5-8B ↓1 79%
Task Understanding
11 models we track have a score here.
- 1 Kimi K2.5 74.8%
- 2
Gemini 3 Flash ↑2 69%
- 3 GLM-5V-Turbo 68.5%
- 4
Gemini 2.5 Flash Sep 2025 ↑3 65.9%
- 5
Seed1.5-VL 64.6%
- 6
Qwen3.5-35B-A3B ↓4 63.6%
- 7
InternVL3-38B ↑1 32.6%
- 8 GLM-4.6V-Flash ↓2 24.8%
- 9
InternVL3.5-30B-A3B ↑1 17.2%
- 10
InternVL3.5-8B ↓1 8.80%
Action Planning
11 models we track have a score here.
- 1 GLM-5V-Turbo ↑2 88%
- 2
Qwen3.5-35B-A3B 82.5%
- 3
Seed1.5-VL ↑2 79%
- 4
Gemini 3 Flash 77%
- 5 Kimi K2.5 ↓4 75.5%
- 6 GLM-4.6V-Flash 74.5%
- 7
InternVL3.5-30B-A3B ↑3 42%
- 8
Gemini 2.5 Flash Sep 2025 ↓1 38.5%
- 9
InternVL3-38B ↓1 37%
- 10
InternVL3.5-8B ↓1 37%
Action Execution
11 models we track have a score here.
- 1 Kimi K2.5 80.5%
- 2
Qwen3.5-35B-A3B 68.5%
- 3 GLM-5V-Turbo 66.5%
- 4
Gemini 2.5 Flash Sep 2025 ↑3 62.5%
- 5
Gemini 3 Flash ↓1 59.5%
- 6 GLM-4.6V-Flash 49.5%
- 7
Seed1.5-VL ↓2 42%
- 8
InternVL3-38B 27%
- 9
InternVL3.5-8B 27%
- 10
InternVL3-8B ↑1 24.5%
Dynamic Adaptation
11 models we track have a score here.
- 1 Kimi K2.5 82%
- 2
Gemini 3 Flash ↑2 79.5%
- 3 GLM-5V-Turbo 77.5%
- 4
Gemini 2.5 Flash Sep 2025 ↑3 75%
- 5
Qwen3.5-35B-A3B ↓3 74%
- 6
Seed1.5-VL ↓1 69%
- 7 GLM-4.6V-Flash ↓1 66%
- 8
InternVL3.5-30B-A3B ↑2 64%
- 9
InternVL3-8B ↑2 58%
- 10
InternVL3-38B ↓2 55.5%
Action Anticipation
11 models we track have a score here.
- 1 Kimi K2.5 68%
- 2
Seed1.5-VL ↑3 62%
- 3 GLM-4.6V-Flash ↑3 61%
- 4
Qwen3.5-35B-A3B ↓2 55%
- 5
Gemini 3 Flash ↓1 49.5%
- 6
Gemini 2.5 Flash Sep 2025 ↑1 47%
- 7 GLM-5V-Turbo ↓4 45%
- 8
InternVL3-38B 42.5%
- 9
InternVL3.5-30B-A3B ↑1 37.5%
- 10
InternVL3-8B ↑1 34.5%
Task Progress Awareness
11 models we track have a score here.
- 1 GLM-5V-Turbo ↑2 74.5%
- 2 Kimi K2.5 ↓1 72.5%
- 3
Qwen3.5-35B-A3B ↓1 70.5%
- 4
Seed1.5-VL ↑1 59%
- 5
InternVL3-38B ↑3 51%
- 6
Gemini 2.5 Flash Sep 2025 ↑1 48.5%
- 7
InternVL3.5-30B-A3B ↑3 46%
- 8
Gemini 3 Flash ↓4 44%
- 9
InternVL3.5-8B 43.5%
- 10 GLM-4.6V-Flash ↓4 40.5%
Task Completion Detection
11 models we track have a score here.
- 1
Qwen3.5-35B-A3B ↑1 82%
- 2
Gemini 2.5 Flash Sep 2025 ↑5 81.5%
- 3
Gemini 3 Flash ↑1 74%
- 4 Kimi K2.5 ↓3 72.5%
- 5
Seed1.5-VL 71.5%
- 6 GLM-4.6V-Flash 71.5%
- 7 GLM-5V-Turbo ↓4 71.5%
- 8
InternVL3-38B 55%
- 9
InternVL3.5-8B 48%
- 10
InternVL3-8B ↑1 47%
Internal State
11 models we track have a score here.
- 1
Gemini 3 Flash ↑3 94%
- 2 Kimi K2.5 ↓1 92%
- 3
Qwen3.5-35B-A3B ↓1 92%
- 4
Gemini 2.5 Flash Sep 2025 ↑3 89%
- 5 GLM-5V-Turbo ↓2 86.5%
- 6 GLM-4.6V-Flash 82.5%
- 7
Seed1.5-VL ↓2 81.5%
- 8
InternVL3-38B 72.5%
- 9
InternVL3.5-30B-A3B ↑1 67%
- 10
InternVL3.5-8B ↓1 63.5%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The UniGUI-Bench 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 UniGUI-Bench 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 7 months ago, in April 2026, for GLM-5V-Turbo. 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 (14 entries).
- Pass Models scored 11 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for GLM-5V-Turbo.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Flash (#60).
- Pass Headroom The best is 80.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on aclanthology.org: 11 rows for models we track and 3 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.