MoonshotAI: Kimi K3

moonshotai/kimi-k3

Context: 1,048,576In: $0.72/1MOut: $13.00/1MUpdated 2026-10-04

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...

Intelligence
43.6
Coding
76.2
Agentic
50.0

Metric profile

Trend over time

Classic evals (curated from model cards)

Moonshot Kimi — manually maintained, may lag releases

MMLU-Pro
78%
GPQA Diamond
74%

Design Arena (ELO by category)

CategoryELOWin rateRank
3d140768%
#4
gamedev137961.8%
#3
codecategories137263.4%
#2
uicomponent136262.3%
#3
dataviz135864%
#3
website134459.9%
#2
fullstack131361.8%
#2
svg129959.4%
#6
webapps128559.1%
#4
python-pptxslides127359.2%
#4
mobileapps126656.9%
#3
androidnative126255.4%
#5
htmlslides125659.6%
#2
agenticgamedev125255%
#3
godotgamedev122348.2%
#14