Kimi K3
A trillion-parameter multimodal model for long-running coding agents, research, and knowledge work.
Language models- Parameters
- 2.8T total / 104B active
- Input
- Text / images / video
- Output
- Text
- Context
- Up to 1M tokens
Build with it
- Long-running software agents
- Research and knowledge work
- Multimodal reasoning
Original model card & license