agents benchmark
AndroidWorld
AndroidWorld evaluates an agent's ability to operate in real Android GUI environments, completing multi-step tasks by perceiving screen content and executing touch/type actions.
Updated Aug 7, 2026
Published models4
Registry coverage4
MetricScore
EvidenceB
AndroidWorld leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 85.3% | 100.0% | 4 | C | |
| 2 | AC | 81.0% | 66.7% | 4 | C | |
| 3 | ZA | 75.7% | 33.3% | 4 | C | |
| 4 | AC | 70.3% | 0.0% | 4 | C |
Score distribution
Top published rows on the benchmark's original scale.
AndroidWorld highlights
The top published results on this benchmark's own scale.
What is AndroidWorld?
Definition and scoring fields from the benchmark registry.
AndroidWorld evaluates an agent's ability to operate in real Android GUI environments, completing multi-step tasks by perceiving screen content and executing touch/type actions.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- AndroidWorld
- Modality
- multimodal
- Primary category
- agents
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- androidworld|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about AndroidWorld.
Which model scores highest on AndroidWorld?
Qwen3.8 Max is currently ranked first with 85.3%.
What does AndroidWorld measure?
AndroidWorld evaluates an agent's ability to operate in real Android GUI environments, completing multi-step tasks by perceiving screen content and executing touch/type actions.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
4 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.