Intelligence IAAug 15, 2026Intelligence IA
Article
Moonshot AI's PerceptionBench shows that no frontier model performs well at visual perception, separate from reasoning, indicating...
A major gap in multimodal AI. The benchmark tests models' ability to 'see' rather than reason.
Redaction Data Cube AISource: The Decoder
01
Brief source
Moonshot AI's PerceptionBench shows that no frontier model performs well at visual perception, separate from reasoning, indicating a major gap in multimodal AI. The benchmark tests models' ability to 'see' rather than reason.