Kimi K3 API

Moonshot's Kimi K3, through an OpenAI-compatible API. 6 providers quote it now, and each request goes to the offer with the lowest estimated charge that meets your rules.

Providers compete for each request. The exchange ranks the offers that meet your rules by estimated charge. It fixes your maximum charge before the first token, and you pay for the tokens that the model uses.
The OpenAI and Anthropic APIs. Use the OpenAI or Anthropic SDK that you have. Set the base URL and your key, and send kimi-k3 as the model.
Your data rules. Send x-liquid-retention: none and only offers whose provider does not keep your content compete for the request.

Facts

Model name: kimi-k3
Maker: Moonshot
Context window: 1M tokens
Input price: $1.200 to $3.000 per million tokens
Output price: $0.100 to $15.000 per million tokens
Providers: 6 independent providers
Regions: asia, cn, eu, global, us

Providers

engy: asia, $1.200 input and $3.000 output per million tokens, $0.120 cache read per million tokens, 1M tokens of context
Pera: asia, $1.800 input and $0.100 output per million tokens, $0.180 cache read per million tokens, $0.00025 fee for each request, 1M tokens of context
deeprelay: cn, $1.950 input and $9.750 output per million tokens, $0.195 cache read per million tokens, 1M tokens of context
MarsCompute: global, $3.000 input and $15.000 output per million tokens, $3.000 cache read per million tokens, 1M tokens of context
EmpirioLabs AI: global, $3.000 input and $15.000 output per million tokens, $3.000 cache read per million tokens, 1M tokens of context
TensorX: eu, $3.000 input and $15.000 output per million tokens, $0.750 cache read per million tokens, 1M tokens of context
Pera: us, $2.600 input and $2.458 output per million tokens, $2.600 cache read per million tokens, $0.02 fee for each request, 1M tokens of context

Scores

HLE: 46.9, #17 of 273
GPQA: 93.5, #6 of 250
AA-LCR: 88.7, #1 of 237
SciCode: 59.5, #8 of 123
SimpleQA Verified: 62.2, #16 of 126

Scores: Artificial Analysis (https://artificialanalysis.ai/)

Scores: Epoch AI, 'Capabilities & Benchmarking' (https://epoch.ai/benchmarks)

Request

curl https://router.inference.ai.exchange/v1/chat/completions \
  -H "Authorization: Bearer $LIQUID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [{"role":"user","content":"Say hello."}]
  }'

Questions

What is Kimi K3?: Kimi K3 served through Grizzly Grid. 1M context window.
How much does Kimi K3 cost?: Offers now ask $1.200 to $3.000 per million input tokens and $0.100 to $15.000 per million output tokens. A job pays for the tokens it uses. The exchange picks the cheapest eligible offer.
What is the context length of Kimi K3?: The largest offer takes 1,048,576 input tokens. One job can generate up to 131,072 tokens. A smaller offer can take less. The exchange skips an offer that cannot take your message.
What can Kimi K3 do?: Its providers list Vision, Video input, Reasoning effort, JSON object, JSON schema, Frequency penalty, Min-p sampling, Presence penalty, Repetition penalty, Seed, Stop sequences, Top-k sampling, Top-p sampling, Text, Tools.
What inputs and outputs does Kimi K3 support?: Text · Images · Video → Text.
Which providers serve Kimi K3?: 6 independent providers quote it now, with 7 offers. The table above ranks them by the cost of a typical job.
How do I call Kimi K3?: Send "kimi-k3" as the model in an OpenAI or Anthropic request. The API section links to setup steps for each client.
When was Kimi K3 listed?: The exchange first listed it on 10/7/2026. This is not its release date.