StudyArena

Kimi K3

FrontierOpen weightsTextReasoning

Kimi family model by Kimi.

About the Kimi family

Moonshot AI's Kimi models: K2 (open ~1T-param MoE, ~32B active), K2 Thinking, K2.5/K2.6/K2.7 Code and K3 with effort tiers, plus Kimi Linear's hybrid attention research. Known for agentic tool use and the Muon optimizer.

All Kimi models

How this one is configured

This configuration runs in reasoning mode (thinks step-by-step before answering).

Also runs at: low · max effort

Effort controls how long the model thinks before answering — higher effort usually means better answers on hard problems, at the cost of speed and money. Same model underneath, so they share this page.

Who makes it

KimiChinaMoonshot AI's brand — open trillion-parameter MoE models.

More from Kimi

← Back to all models