Moonshot AI

Kimi models, open-weight agentic performance at low cost

moonshot.ai platform.moonshot.ai/docs

Overall in llm apis

13of 18 · score 59.4

Data coverage

85%

23 of 27 fields sourced

Best for

•Agent loops on an open-weight budget
•Self-hosting a tool-use-capable model under a permissive licence
•Replacing a frontier model on the cheaper half of a mixed workload

Strong on

+Trains on your data Partial 2/18

Weak on

−Zero retention No 16/18
Better: Amazon Bedrock: Yes
−Audio in/out No 11/18
Better: OpenAI: Partial
−$/M output $15/M tok 10/13
Better: Fireworks AI: $0.05/M tok
−Batch discount 40% 10/14
Better: Anthropic: 50%

vs Qwen

•Ahead on Image input: Yes vs Partial
•Ahead on Open weights: Yes vs Partial
•Behind on Zero retention: No vs Partial
•Behind on MCP support: No vs Partial

How Moonshot AI scores among 18 llm apis23 of 27 fields known

Capability7

Flagship model

Kimi K3 (kimi-k3)

platform.kimi.ai · 2026-07-30

Context window1st

1,048,576 tokens

1/14

platform.kimi.ai · 2026-07-30

Max output

—

Image input1st

Yes

1/18

platform.kimi.ai · 2026-07-30

Audio in/out

11/18

Leader: OpenAI · Partial

· 2026-01-15

Open weights1st

Yes

1/18

github.com · 2026-01-15

OpenAI-compat API1st

Yes

1/18

platform.kimi.ai · 2026-07-30

Performance2

TTFT p50

—

Output tok/s

—

Pricing4

$/M input

$3/M tok

10/13

Leader: Groq · $0.05/M tok

platform.kimi.ai · 2026-07-30

$/M output

$15/M tok

10/13

Leader: Fireworks AI · $0.05/M tok

platform.kimi.ai · 2026-07-30

$/M cache read

$0.3/M tok

7/12

Leader: DeepSeek · $0.0028/M tok

platform.kimi.ai · 2026-07-30

Batch discount

40%

10/14

Leader: Anthropic · 50%

platform.kimi.ai · 2026-11-07

Agentic6

Tool use1st

Yes

1/18

platform.kimi.ai · 2026-07-30

Schema output

Partial

11/18

Leader: Anthropic · Yes

· 2026-01-15

Effort control

Partial

5/18

Leader: Anthropic · Yes

· 2026-01-15

Computer use

5/18

Leader: Anthropic · Yes

· 2026-01-15

MCP support

6/18

Leader: Anthropic · Yes

· 2026-01-15

Cache TTL

Automatic, no fixed TTL

platform.kimi.ai · 2026-07-30

Governance & continuity6

Continuity policy

Dated model IDs (the -0905 style suffix) are used and remain addressable, which is better than DeepSeek's rolling aliases. No published deprecation policy or notice window. As with the other open-weight labs, the real guarantee is the licence, download the checkpoint and continuity is your problem, not theirs.

· 2026-01-15

Notice period

—

Pinnable versions1st

Yes

1/18

· 2026-01-15

Zero retention

16/18

Leader: Amazon Bedrock · Yes

· 2026-01-15

Trains on your data2nd