Between Kimi K3 and DeepSeek V4: The Time Gap of Native Multimodality
The article discusses the strategic importance of native multimodal capabilities in AI models, using Kimi K3 as a case study. It highlights how visual feedback is becoming crucial for agent tasks, and contrasts different approaches among Chinese AI labs regarding when and how to integrate vision into large language models.
Moonshot AI
DeepSeek
OpenAI
Alibaba/Qwen
ByteDance
Tencent
Apple
36Kr04.08 · 10:02

