moonshotaiCertifiedReleased 07/16/2026

Kimi K3

Kimi’s flagship model for long-horizon coding and end-to-end knowledge work, with a 1M-token context window.

View API reference ↓
Input / output priceFrom$0.18 / $10.00per 1M
24h uptime99.98%
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.inferencebureau.com/v1",
  apiKey: process.env.IB_API_KEY,
});

const res = await client.chat.completions.create({
  model: "moonshotai/kimi-k3",
  messages: [{ role: "user", content: "Why is the sky blue?" }],
});

console.log(res.choices[0].message.content);

Providers

Dispatch routes requests across these providers and fails over between them without changing the model. Restrict or order providers with team Directives. Prices are gateway list prices before the platform fee.

ProviderContextMax outputLatencyThroughputInputOutputCacheCapabilitiesZDRNo trainingRegionsStatus
alibaba1M1M1.3 s55 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
Amazon Bedrock1M1M4.6 s111 tps$3.00$15.00Read $0.30Write $3.757 featuresYesYesUSCertified
Baseten1M131K0.9 s73 tps$3.00$15.00Read $0.30Write —7 featuresYesYesUSCertified
boundless1M1M0.4 s236 tps$2.30$11.40Read $0.23Write —6 featuresNoNoGlobalCertified
DeepInfra1.05M131K0.4 s68 tps$2.85$14.25Read $0.29Write —7 featuresNoYesGlobalCertified
digitalocean1M131K0.6 s53 tps$2.85$14.25Read $0.29Write —7 featuresYesYesGlobalCertified
Fireworks1M131K0.6 s61 tps$3.00$15.00Read $0.30Write —7 featuresYesYesUSCertified
inference-net1.05M1.05M0.4 s—$3.00$15.00Read $0.30Write —6 featuresNoNoGlobalCertified
modal1M1M0.6 s145 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
moonshotai1M131K1.6 s51 tps$3.00$15.00Read $0.30Write $3.008 featuresYesYesGlobalCertified
morph1M1M0.5 s82 tps$2.17$12.00Read $0.53Write —7 featuresYesYesGlobalCertified
nebius1.02M131K0.4 s182 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
parasail1M1M0.3 s171 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
relace1.05M1.05M1.3 s129 tps$0.18$11.01Read $0.18Write —6 featuresYesYesGlobalCertified
runware1.05M1.05M0.4 s226 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
sail-research1.05M1.05M0.5 s81 tps$1.50$10.00Read $0.20Write —6 featuresNoNoGlobalListed
Together AI1M131K0.5 s114 tps$3.00$15.00Read $0.30Write —7 featuresYesYesGlobalCertified
wafer1M131K0.3 s—$0.72$16.00Read $0.71Write —7 featuresYesYesUSCertified

Uptime

24 hours

Request success rate per provider.

alibaba100.00%
Amazon Bedrock99.83%
Baseten99.92%
boundless100.00%
DeepInfra100.00%
digitalocean99.97%
Fireworks100.00%
inference-net100.00%
modal100.00%
moonshotai100.00%
morph100.00%
nebius100.00%
parasail100.00%
relace99.97%
runware100.00%
sail-research100.00%
Together AI100.00%
wafer100.00%

Throughput

last hour

P50 output tokens per second.

boundless236 tps
runware226 tps
nebius182 tps
parasail171 tps
modal145 tps
relace129 tps
Together AI114 tps
Amazon Bedrock111 tps
morph82 tps
sail-research81 tps
Baseten73 tps
DeepInfra68 tps
Fireworks61 tps
alibaba55 tps
digitalocean53 tps
moonshotai51 tps
inference-net—
wafer—

Latency

last hour

P50 time to first token.

wafer0.3 s
parasail0.3 s
runware0.4 s
inference-net0.4 s
nebius0.4 s
boundless0.4 s
DeepInfra0.4 s
Together AI0.5 s
morph0.5 s
sail-research0.5 s
modal0.6 s
Fireworks0.6 s
digitalocean0.6 s
Baseten0.9 s
alibaba1.3 s
relace1.3 s
moonshotai1.6 s
Amazon Bedrock4.6 s

Getting started

Create a key on the API Keys page and set it as IB_API_KEY. See the quickstart.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.inferencebureau.com/v1",
  apiKey: process.env.IB_API_KEY,
});

const res = await client.chat.completions.create({
  model: "moonshotai/kimi-k3",
  messages: [{ role: "user", content: "Why is the sky blue?" }],
});

console.log(res.choices[0].message.content);

Top-level parameters

Standard Chat Completions parameters work as documented. These have model-specific behavior.

ParameterTypeRequiredDescription
modelstringYesModel ID in the form creator/model, e.g. moonshotai/kimi-k3. Must be allowed by your team's Directives.
max_tokensnumberNoHard cap on generated tokens, and the basis for the balance reserved per request. Kimi K3 supports up to 131,072 output tokens.
reasoning{ effort: 'low' | 'medium' | 'high' }NoReasoning effort, mapped to the provider's native configuration. Reasoning tokens count toward output usage.
toolsTool[]NoFunction definitions the model can call. Tool calls are returned in choices[].message.tool_calls.
Supportedmax_tokensstoptoolstool_choicereasoninginclude_reasoningresponse_formatstructured_outputs

Input limits

InputSourcesLimits
Text—Prompt and response share the 1M-token context window
ImageURL, base64Sent as image_url content parts; counts as input tokens
PDFURL, base64Sent as file content parts; counts as input tokens

More models by moonshotai

ModelContextInputOutputCacheRelease
moonshotai/kimi-k3.1-pro1M$2.00$10.00Read $0.20Write —10/23/2026
moonshotai/kimi-k3.1-pro-highspeed1.05M$4.00$20.00Read $0.20Write $5.0010/23/2026
moonshotai/kimi-k3-fast1M$6.00$22.50Read $0.60Write —07/27/2026
moonshotai/kimi-k2.7-code-highspeed262K$1.90$8.00Read $0.38Write —06/15/2026
moonshotai/kimi-k2.7-code256K$0.95$4.00Read $0.19Write —06/12/2026
moonshotai/kimi-k2.6262K$0.95$4.00Read $0.16Write —04/20/2026
moonshotai/kimi-k2.5256K$0.60$3.00Read —Write —01/26/2026
moonshotai/kimi-k2131K$0.57$2.30Read —Write —07/11/2025