AGI Research Team sets new record on AndroidWorld benchmark
HIGHLIGHTS
- •AGI Research Team achieved 97.4% task success on AndroidWorld benchmark, surpassing all previous systems
- •AGI-0 Agent operates from screenshots using a multimodal vision-language model fine-tuned for UI understanding, without relying on accessibility trees or hidden UI metadata
- •This demonstrates AGI-0's ability to perform tasks on mobile devices like a human, setting a new state-of-the-art performance record
Published:
Official Source (Official Blog)