General Agents Benchmark

AndroidWorld

AndroidWorld - visual agent benchmark for Android device task completion.

Category

General Agents

Score type

percent

Scale

100%

higher is better

Models tested

8

What is AndroidWorld?

AndroidWorld is a benchmark for how well a model acts as a general-purpose agent. The official description says: "AndroidWorld - visual agent benchmark for Android device task completion". Below we describe in simple words what benchmarks of this type generally examine.

What does it test?

AndroidWorld measures how well a model acts as a general-purpose agent. In detail, it checks:

  • solving long, multi-step tasks with many intermediate decisions
  • using tools correctly (browsers, terminals, APIs, calendars, other apps)
  • adapting when something unexpected happens mid-task
  • staying on goal over long task horizons instead of wandering off

How to read the score

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

Official methodology note: AndroidWorld - visual agent benchmark for Android device task completion.

Frequently asked questions

What is the AndroidWorld benchmark?

AndroidWorld is a benchmark for how well a model acts as a general-purpose agent. The official description says: 'AndroidWorld - visual agent benchmark for Android device task completion'. Below we describe in simple words what benchmarks of this type generally examine.

What does AndroidWorld test?

AndroidWorld mainly measures how well a model acts as a general-purpose agent. Concretely, it checks: solving long, multi-step tasks with many intermediate decisions; using tools correctly (browsers, terminals, APIs, calendars, other apps); adapting when something unexpected happens mid-task.

How is the AndroidWorld score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get AndroidWorld data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/androidworld/scores) or pull it directly from the Open LLM Radar API with any HTTP client.

Model scores 8 models tested

Reset
# Model Score Date Source Verified
1 Qwen3.8-Flash-Next
Qwen Team (Alibaba)
84.5%
26.08.2026 Report Verified
2 Qwen3.8-27B
Qwen Team (Alibaba Cloud)
81.9%
18.08.2026 Report Verified
3 Qwen3.5-35B-A3B
Qwen Team (Alibaba Cloud)
71.1%
01.02.2026 Report Unverified
4 Qwen3.6-27B
Qwen Team (Alibaba Cloud)
70.3%
22.04.2026 Qwen Team (official model card) Verified
5 Qwen3.5-122B-A10B
Qwen Team (Alibaba Cloud)
66.4%
01.02.2026 Report Verified
6 Qwen3.5-27B
Qwen Team (Alibaba Cloud)
64.2%
22.04.2026 Qwen Team (official model card - comparison) Unverified
7 Qwen3-235B-A22B
Qwen Team (Alibaba Cloud)
63.7%
01.02.2026 Report Unverified
8 Qwen3.5-4B
Qwen
58.6%
01.02.2026 Report Verified

What now?

You've seen how models score on AndroidWorld — here's what to do next.