Cheese Dossier · CF-1.3

Gemini 2.5 Pro

google · Cheese #155· 6 signals · specialty gpqa · Proprietary

CHEESE INDEX
43.2
DEPTH
52.9
BALANCE
77.5
CONSENSUS
68.2
STABILITY
84.8
COVERAGE
77%
VALUE
3.84
I / O
$1.25 / $10
SPECIALTY
gpqa +8
BEST FIT
long_horizon 41

深度雷達

codingagentmathknowleinstrumultimmultilreason

實線=本模型 · 虛線=群體中位

芝士軸

coding25.3
agent47.8
horizon32.1

Cheese Index = 可用軸中位數;Balance = 三軸分歧小。

能力細節

swe63.2
gpqa ★83.0
hle17.8
coding37.8
agent47.8
math35.2
knowledge29.4
instruction57.9
multimodal ★70.3
reasoning68.4

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 Gemini 3.5 Flash-Lite

coding-4.4
agent+4.7
horizon-4.0