GUI Task Completion on AndroidWorld
57.8Success Rate (SR)MIRAGE-8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MIRAGE-8BSize=8B2026.06 | 57.8 | 13.7 | 27 | |
| MIRAGE-4BSize=4B2026.06 | 52.6 | 14.2 | 31 | |
| Qwen3-VL-8B-InstructSize=8B2026.06 | 47.6 | 12.6 | 108 | |
| MAI-UI-2BSize=2B2026.06 | 46.5 | 15 | 116 | |
| Qwen3-VL-4B-InstructSize=4B2026.06 | 42.9 | 14.3 | 103 | |
| UI-Venus-Navi-7BSize=7B2026.06 | 42.2 | 13.3 | 92 | |
| UI-TARS-7B-SFTSize=7B2026.06 | 33.6 | 17.3 | 104 | |
| UI-TARSSize=7B2026.06 | 33 | — | — | |
| GPT-4oSize=–2026.06 | 30.6 | — | — | |
| Ferret-UI LiteSize=3B2026.06 | 28 | — | — | |
| GUICrafter-3BTraining Stage=Full (Stage 1+2)2026.06 | 25.43 | — | — | |
| GUICrafter-3BTraining Stage=only Stage12026.06 | 14.66 | — | — | |
| GUI-R1-3B2026.06 | 14.22 | — | — | |
| Qwen2.5-VL-3B2026.06 | 10.34 | — | — | |
| ShowUI-2BSize=2B2026.06 | 7.8 | 18.9 | 33 |