AgentWorldBench Android
AgentWorldBench Android domain - open-ended rubric evaluation for language world models simulating Android device environments. Normalized to 0-100.
Category
General Agents
Max Score
100.0
Score Type
percent
Active
Yes
Model Scores
Scores for AgentWorldBench Android
| Model | Score | Date | Verified |
|---|---|---|---|
| Qwen-AgentWorld-397B-A17B |
83.51%
|
24.06.2026 | Unverified |
| Qwen3.5-35B-A3B |
8.35%
|
24.06.2026 | Unverified |
| Qwen3.6-Plus |
56.21%
|
24.06.2026 | Unverified |
| GPT-5.4 |
81.37%
|
24.06.2026 | Unverified |
| Claude Opus 4.8 |
97.43%
|
24.06.2026 | Unverified |
| Claude Opus 4.6 |
100.00%
|
24.06.2026 | Unverified |
| Claude Sonnet 4.6 |
60.28%
|
24.06.2026 | Unverified |
| DeepSeek-V4-Pro |
29.66%
|
24.06.2026 | Unverified |
| GLM-5.1 |
71.73%
|
24.06.2026 | Unverified |
| Kimi K2.6 |
69.91%
|
24.06.2026 | Unverified |
| MiniMax-M2.7 |
52.40
|
24.06.2026 | Unverified |
| Qwen-AgentWorld-35B-A3B |
61.78%
|
24.06.2026 | Verified |
| Qwen3.5-397B-A17B |
26.77%
|
24.06.2026 | Unverified |
| Gemini 3.1 Pro |
96.36%
|
24.06.2026 | Unverified |