Kimi K1.5
K1.5 combined long context with RL-based reasoning and provided a research base for later open Kimi agent models.
FAMILY / MOONSHOT AI
Kimi
A family that grew from long-context and RL research into large open MoE and native visual agents.
- First model
- 2025
- Models
- 6
- Latest
- Kimi K3
01 / LINEAGE MAP
Model lineage
predecessor derived from
02 / GENERATIONS
What changed
Kimi K2
K2 paired trillion-scale MoE with an open agent model and demonstrated large-scale training with the Muon optimizer.
Kimi K2 Thinking
A K2-based thinking agent trained to reason while using tools natively.
Kimi K2.5
A natively multimodal continuation of K2 with visual coding and an Agent Swarm of up to 100 subagents.
Kimi K2.6
The open successor to K2.5 with stronger long-horizon coding, general agents, vision, and Agent Swarm.
Kimi K3
K3 pushed open-model scale toward 3T parameters while unifying agent, vision, and long-context capabilities.