Cheese Dossier · CF-1.3
Kimi K2
Moonshot AI · Cheese #166· 2 signals · specialty knowledge · Proprietary
CHEESE INDEX
42.3
DEPTH
51.1
BALANCE
—
CONSENSUS
76.5
STABILITY
100.0
COVERAGE
31%
VALUE
13.63
I / O
$0.6 / $2.5
SPECIALTY
knowledge +-2
BEST FIT
budget 40.7
深度雷達
實線＝本模型 · 虛線＝群體中位
芝士軸
horizon
Cheese Index = 可用軸中位數；Balance = 三軸分歧小。
能力細節
math
knowledge
instruction
reasoning
部署建議
中段指數：適合成本敏感路徑。優先看 Pareto／Value；用深度雷達找專長匹配，再與更高指數模型 A/B。
鄰近差分
vs DeepSeek V3.1
coding—
horizon—
指數鄰近
- DeepSeek V3.1 42.3
- Gemini 3 Pro Deep Think 42.4
- Mistral Medium 3.5 128B 42.0
- GPT-5.4 nano 41.9
- Claude Haiku 4.5 42.9
同組織
- Kimi K2.5 (Reasoning) 49.9
- Kimi K2.7 Code 48.0
- Kimi K2.5 34.0