MoonshotAI: Kimi K3

moonshotai/kimi-k3

Kimi K3 is MoonshotAI's open-weight flagship and the strongest model in its open-weight class, ranking #17 of 263 on the TryAii Score. It's strongest on head-to-head chat (#13 on Chatbot Arena Elo (Code)). API pricing starts at $0.239/$11.94 per million input/output tokens, though similar scores now sell for about $8 per million output tokens, with 18 independent hosts serving it. It's reasonably quick at roughly 41-65 tokens/sec.

Prompt $/1M

$0.239

Completion $/1M

$11.936

Context

1,048,576 tok

Best provider

94 tok/s · InferenceNet

Specs

Model id: moonshotai/kimi-k3
Display name: MoonshotAI: Kimi K3
Modality: text+image+video->text
Tokenizer: Other
Instruct type: —
Context length: 1,048,576
Max completion: 943,718
Prompt $/1M: $0.239
Completion $/1M: $11.936
Free tier: no

Capabilities

Function callingno
JSON modeno
System promptno
Visionyes
Moderatedno

Benchmark scores

Each axis is one benchmark, normalized to the best model in the field.

No benchmark data for this model.

Speed by provider

Provider	tok/s	TTFT	Total latency
InferenceNetfastest	94	693 ms	885 ms
Fireworks	79	1,172 ms	1,511 ms
Wafer	77	786 ms	1,019 ms
Parasail	69	823 ms	917 ms
Modal	66	1,021 ms	1,550 ms
BaseTen	65	2,543 ms	3,679 ms
Phala	63	1,464 ms	1,862 ms
Morph	62	1,722 ms	2,532 ms
Sail Research	60	3,277 ms	4,771 ms
Relace	54	1,873 ms	2,330 ms
Decart	52	4,154 ms	5,452 ms
Makora	49	1,677 ms	2,231 ms
Together	47	1,097 ms	1,563 ms
AkashML	46	1,583 ms	2,441 ms
Amazon Bedrock	45	5,194 ms	6,594 ms
DigitalOcean	41	1,541 ms	1,925 ms
Artificial Analysis	41	3,103 ms	—
Alibaba	31	3,216 ms	4,353 ms
Moonshot AI	19	6,905 ms	7,889 ms
DeepInfra	16	2,206 ms	3,063 ms
Chutes	13	4,792 ms	6,079 ms