Moonshot AILanguage modelSupersededOpen weights
Kimi K2
The trillion-parameter open model that kicked off the 'Kimi moment'.
Overall Intelligence
#48
of 56
25.0 pts behind #1
0
races led
0
weeks at #1
63.7
score
- Released
- Jul 11, 2025
- 1.3 yr ago
- Context
- 128K
- 16K max output
- Price
- $0.6 / $2.5
- per 1M in/out · $1.07 blended
- Speed
- 50 tok/s
- 0.8s to first token
- Modalities
- text
- outputs text
- Weights
- Open
- 1T MoE (32B active)
The 30-second version
Moonshot's trillion-parameter open model that made Chinese open weights a serious agentic coding option in mid-2025.
Key takeaways
- 1Open weights at trillion-parameter scale.
- 2Strong tool use for an open model.
Where it ranks
Current snapshot · illustrative scores · click any race to watch it
Kimi K2Best on each axis
Overall
Coding
48Coding+25.6 pts49Frontend & UI+155 Elo48Backend & APIs+22.8 pts48Distributed Systems+26.8 pts48Full-Stack Apps+23.6 pts48Debugging & Repair+25.8 pts48Agentic Coding+26.1 pts47Code Review & Security+23.1 pts51Data Science & SQL+25.8 pts47Mobile Apps+21.8 pts46Terminal & DevOps+28.4 pts50Competitive Programming+771
Hard Reasoning
Math
Science
Writing & Language
Multimodal
Tool Use & Agents
Overall Intelligence over time
#1 frontierKimi K2V4.1 FlashGLM-5.3
Update history
The distilled version first; open any release for the deeper breakdown.
The Kimi family