AnyAppBench
Agents repeat the same task in different Android apps, with and without suggested steps. Each app has a verifier that checks the same success condition.
The leaderboard compares 13 agents on 520 task and app combinations across 52 apps. Of 100 task templates, 48 come from AndroidWorld.
- Made by
- KAIST and Carnegie Mellon University
- Released
- September 2026
- Kind
- Phones
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Phone apps
- Items from
- Another benchmark, Generated
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- AndroidWorld
- In views
- Computer use, Agents
- Strongest model tested
-
Gemini 3.1 Pro #40 on the capability index
- Last model added
- February 2026
- Scores refreshed
- November 20th, 2026
Hybrid with UI-Venus-Ground-72B (Gemini 3.1 Pro + UI-Venus-Ground-72B) tops the leaderboard with 67.9% Gemini 3.1 Pro leads with 67.8% among models released by February 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran on its own with several settings, we keep its best. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. The best model on its own is Gemini 3.1 Pro, at 67.8%. Agents built around a model, such as Hybrid with UI-Venus-Ground-72B (Gemini 3.1 Pro + UI-Venus-Ground-72B), are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.
The newest model with a score here came out in February 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
- 67.9 Hybrid with UI-Venus-Ground-72B (Gemini 3.1 Pro + UI-Venus-Ground-72B)
- 67.8
Gemini 3.1 Pro
- 59.0 Hybrid with UI-Venus-Ground-72B (GPT-5.1 + UI-Venus-Ground-72B)
- 54.1 GPT-5.1
- 36.3 Mobile-Agent-v3
- 35.0 MAI-UI-8B
- 32.8
Qwen2.5-VL-72B
- 31.2 UI Voyager-4B
- 31.2 UI-Venus-Navi-72B
- 25.6 UI-Venus-7B
- 23.0 Qwen3-VL-30B-A3B
- 22.4
GUI-Owl-7B
- 18.9 Qwen3-VL-8B
- 1 Hybrid with UI-Venus-Ground-72B (Gemini 3.1 Pro + UI-Venus-Ground-72B) 67.9%
- 2
Gemini 3.1 Pro 67.8%
- 3 Hybrid with UI-Venus-Ground-72B (GPT-5.1 + UI-Venus-Ground-72B) 59%
- 4 GPT-5.1 54.1%
- 5 Mobile-Agent-v3 36.3%
- 6 MAI-UI-8B 35%
- 7
Qwen2.5-VL-72B 32.8%
- 8 UI Voyager-4B 31.2%
- 9 UI-Venus-Navi-72B 31.2%
- 10 UI-Venus-7B 25.6%
- 11 Qwen3-VL-30B-A3B 23%
- 12
GUI-Owl-7B 22.4%
- 13 Qwen3-VL-8B 18.9%
Source: Official AnyAppBench leaderboard
- 1
Gemini 3.1 Pro 67.8%
- 2 GPT-5.1 54.1%
- 3
Qwen2.5-VL-72B 32.8%
- 4
GUI-Owl-7B 22.4%
Source: Official AnyAppBench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Hybrid with UI-Venus-Ground-72B (Gemini 3.1 Pro + UI-Venus-Ground-72B) | 67.9% | Feb 2026 | Official leaderboard | |
| 2 | | 67.8% | Feb 2026 | Official leaderboard | |
| 3 | Hybrid with UI-Venus-Ground-72B (GPT-5.1 + UI-Venus-Ground-72B) | 59% | Nov 2025 | Official leaderboard | |
| 4 | GPT-5.1 | 54.1% | Nov 2025 | Official leaderboard | |
| 5 | Mobile-Agent-v3 | 36.3% | Official leaderboard | ||
| 6 | MAI-UI-8B | 35% | Official leaderboard | ||
| 7 | | 32.8% | Jan 2025 | Official leaderboard | |
| 8 | UI Voyager-4B | 31.2% | Official leaderboard | ||
| 9 | UI-Venus-Navi-72B | 31.2% | Official leaderboard | ||
| 10 | UI-Venus-7B | 25.6% | Official leaderboard | ||
| 11 | Qwen3-VL-30B-A3B | 23% | Official leaderboard | ||
| 12 | | 22.4% | Aug 2025 | Official leaderboard | |
| 13 | Qwen3-VL-8B | 18.9% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 67.8% | Feb 2026 | Official leaderboard | |
| 2 | GPT-5.1 | 54.1% | Nov 2025 | Official leaderboard | |
| 3 | | 32.8% | Jan 2025 | Official leaderboard | |
| 4 | | 22.4% | Aug 2025 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard in 2 groups: aw-inherited templates and new anyappbench templates. Each ranks models on one part of the test. Gemini 3.1 Pro leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
AW app
4 models we track have a score here.
- 1
Gemini 3.1 Pro 81.1%
- 2 GPT-5.1 60.4%
- 3
Qwen2.5-VL-72B 38.4%
- 4
GUI-Owl-7B 31.6%
New apps
4 models we track have a score here.
- 1
Gemini 3.1 Pro 52.1%
- 2 GPT-5.1 41.8%
- 3
Qwen2.5-VL-72B 25.3%
- 4
GUI-Owl-7B 19.5%
AW app
4 models we track have a score here.
- 1
Gemini 3.1 Pro 79%
- 2 GPT-5.1 66.6%
- 3
Qwen2.5-VL-72B 40%
- 4
GUI-Owl-7B 21.6%
New apps
4 models we track have a score here.
- 1
Gemini 3.1 Pro 58.9%
- 2 GPT-5.1 47.8%
- 3
Qwen2.5-VL-72B 27.4%
- 4
GUI-Owl-7B 16.7%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official AnyAppBench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official AnyAppBench 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 9 months ago, in February 2026, for Gemini 3.1 Pro. 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 official leaderboard (13 entries).
- Pass Models scored 4 models we track, plus 9 other entries on its leaderboard.
- Warning Newest score The newest came 9 months ago, in February 2026, for Gemini 3.1 Pro.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3.1 Pro (#40).
- Pass Headroom The best is 67.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on anyappbench.github.io: 6 rows for models we track and 7 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.