Kimi K2.5 covers MMMU Pro and VideoMMMU: can visual understanding and video inference be stable?

Kimi K2.5 Open Source Multimodal Agent Full Solution: Parallel Collaboration with Visual Programming and Agent Swarm

1. Abstract Kimi K2.5 is an open-source "vision + agentic" multimodal model released by Moonshot AI, which supports unified image/video and text input...