Kimi K3

moonshotai · moonshotai/kimi-k3

Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work

≈699k Chinese characters of context · Reads images · Step-by-step reasoning · Calls tools / agents · Structured JSON output · Open weights, self-hostable

Read a 200,000-Chinese-character book and write a 2,000-Chinese-character summary: about $0.94

An order-of-magnitude estimate at 1.5 tokens per Chinese character and the vendor's own rate — not a quote. Full method: the glossary on the AI Models page.

Specs

Context1,048,576(≈699k Chinese characters)
Max output131,072
Structured outputYes
Open weightsYes
Released2026-07-16

Source: models.dev snapshot 2026-08-29

Who serves it, at what price

ProviderIn /MOut /MCache read /M
Moonshot AIfirst-party$3$15$0.3
AIHubMix$3$15$0.3
CoralBricks$3$15$0
Cortecs$3$14.999
CrofAI$2$8$0.25
DigitalOcean$2.85$14.25$0.285
EmpirioLabs AI$3$15$3
GitHub Copilot$3$15$0.3
GreenPT$3.762$18.81$0.9405
Charm Hyper$3.2664$16.332$0.32664
KenariBilled by planBilled by plan
DevPass (LLM Gateway)$2.83$14.13$0.28
Moonshot AI (China)$3$15$0.3
Neon$3$15$0.3
Neuralwatt$3$15$0.3
Ollama Cloud
OpenCode Zen$3$15$0.3
OpenCode Go$3$15$0.3
Requesty$2.25$11.25$0.225
Tinfoil$4$20$0.8
Vancine$2.4$12$0.24
Venice AI$3.75$18.75$0.375
Vivgrid$3$15$0.3

Rows tagged “first-party” are the model vendor's own pricing; untagged rows are resale channels and may differ.

When to choose it

  • 需要百万级上下文一次性吃进长文档,且不想自己做分块检索
  • 开源权重可自部署,对数据出境有顾虑时是少数可选项

Caveats

  • 开源权重不等于自部署便宜 —— 这个体量的模型自己跑,硬件成本通常高于直接调 API

Basis: first-party pricing page · reviewed 2026-08-09

Leaderboard

artificial-analysisIntelligence Index v4.1.1

Snapshot: 2026-08-09