Moonshot AI

Available

Kimi K3 on NexoRouter

API ID: kimi-k3

kimi-k3 is Moonshot's flagship thinking Chat Completions ID on NexoRouter. Choose it for long-context agent work when you can budget for reasoning tokens and you will send the exact catalog string kimi-k3.

Quick start Try in Playground Official source

Type: Chat
Input / generation: $3.04 / 1M tokens
Output: $15.20 / 1M tokens
Endpoint: /v1/chat/completions

Verified information

Overview

kimi-k3 is presented as a chat model from Moonshot AI. Start with an OpenAI-compatible call and evaluate it against your product's prompts, language, and quality criteria.

Verified routePublished endpoint

Publication confirms the /v1/chat/completions route; it does not claim features the contract has not documented.

Commercial statusReviewed price

The displayed price is NexoRouter's approved public price, not an internal cost or supplier discount.

Profile boundaryNo invented metrics

Context, tools, benchmarks, latency, TTFT, and availability appear only when they have a source and measurement window.

Editorial review

Reviewed technical profile

When to choose this model

Use kimi-k3 when you need Moonshot's current flagship ID for agent loops, long-context analysis, or LatAm apps already on the OpenAI SDK that only need base_url plus model swaps.

Moonshot documents that Kimi K3 always reasons and exposes top-level reasoning_effort with values low, high, and max (default max). Reasoning tokens usually bill as output — set an explicit max_tokens budget instead of a toy ceiling.

Vendor materials state kimi-k2.5 and moonshot-v1 retired on 31 August 2026. Replace those strings in client code after confirming kimi-k3 remains Available on the NexoRouter card.

Flagship thinking model ID on the Moonshot mapping
Plan output budget for reasoning plus final answer
Replace retired kimi-k2.5 / moonshot-v1 strings after catalog check

How to call it on NexoRouter

Use the OpenAI-compatible Chat Completions surface at https://api.nexorouter.com/v1 with Authorization Bearer and your NexoRouter key.

Set the model field to the exact API ID kimi-k3. Marketing names and vendor self-reported aliases are not proof that NexoRouter will route the request.

Use the Integration examples on this page with your NexoRouter API key and the exact model ID above. Point the OpenAI SDK at https://api.nexorouter.com/v1, send Authorization Bearer, and copy the public_id from this card into the model field.

Base URL: https://api.nexorouter.com/v1
Exact model ID: kimi-k3
Endpoint path on the profile: /v1/chat/completions

Pricing and billing units

Current prepaid unit prices appear in the live pricing panel on this model page (input, cached input, and output). That panel is the authority for budgeting and invoices.

NexoRouter bills in prepaid USD per 1M tokens. Do not treat dated blog tables or vendor first-party price sheets as the live contract rate — only the live pricing panel on this page is authoritative.

Authority: live pricing panel on this page
Unit: prepaid USD per 1M tokens
Do not budget from dated blog tables

Practical limits

Moonshot documents that kimi-k3 always reasons and exposes top-level reasoning_effort with values low, high, and max (default max). Confirm your NexoRouter path forwards those fields before depending on them in multi-turn tool loops.

Reasoning tokens usually bill as output — set an explicit max_tokens budget that covers reasoning plus the final answer, not a toy ceiling.

Stick to the verified public Chat Completions surface on this page. If a vendor sample shows a parameter that is absent here, treat it as Moonshot-direct guidance until the catalog card lists it.

Plan max_tokens for reasoning plus final answer
Confirm reasoning_effort / reasoning_content forwarding on your path
Use only parameters shown on this NexoRouter card

Published Sep 9, 2026 UTC after technical, language, and source review. It is shown only for this exact model version and locale.

Delivery contract

Provider & route

kimi-k3 is exposed by NexoRouter under the verified Moonshot AI catalog mapping below. This is the delivery contract customers should use, not a promise of additional upstream routes.

ProviderModel IDEndpointStatus

Moonshot AIkimi-k3/v1/chat/completionsAvailable

Provider names describe the reviewed catalog mapping. NexoRouter does not claim multi-provider failover for a model unless it is measured and published separately.

Public billing

Pricing

Public usage pricing for this model ID. The unit changes by endpoint: not every model is billed by tokens.

Input / generation: $3.04 / 1M tokens
Cached input: $0.30 / 1M tokens
Output: $15.20 / 1M tokens

Cost estimator

Estimate a request with the current public price for this exact model ID.

Input tokensOutput tokens

Estimated total$0.01

Estimate only. Actual usage, rounding, cached tokens, media parameters, and the final billing ledger determine the charged amount.

Prices are published after commercial review. A supplier update creates a review; it never exposes supplier discounts, orders, or internal costs.

Measured, not estimated

Performance

Performance metrics are published only after NexoRouter has enough model-specific production observations and a defined measurement window.

ThroughputNot enough verified data

Tokens per second require streaming observations for this exact model ID.

End-to-end latencyNot enough verified data

A gateway response time is not the same as model inference latency, so the two are not mixed.

Time to first tokenNot enough verified data

TTFT appears only when streaming timestamps are available for a representative sample.

No benchmark or chart is synthesized from another provider. These fields will populate only from NexoRouter measurements with a sample size and time window.

NexoRouter aggregate telemetry

Activity

Privacy-preserving, aggregate model activity from NexoRouter gateway logs. No prompts, API keys, user identities, or IP addresses are exposed.

7-day attempts: 135
7-day tokens: 3.7M
7-day success rate: 98.5%
Catalog token share: 9.6%

Jul 17

Jul 24

Jul 31

Aug 7

Aug 14

Aug 21

Aug 28

Sep 4

Success rate is computed from NexoRouter gateway outcomes for this model ID. It is not an upstream SLA, quality score, or benchmark.

OpenAI-compatible

Integration

NexoRouter exposes this model through /v1/chat/completions. Use a NexoRouter API key and retain the OpenAI-compatible base URL.

curl --fail-with-body 'https://api.nexorouter.com/v1/chat/completions' \
  -H "Authorization: Bearer $NEXOROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kimi-k3",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'

Chat Completions example using the exact model ID. Adjust messages and parameters for your use case.

Start with the API reference

Confirm authentication, endpoint, request shape, and limits before copying an integration into production.

Open guide

Provenance

Sources & updates

Every profile combines NexoRouter operating data with provider documentation links. A provider link does not become a capability promise until that claim is reviewed and published.

NexoRouter public catalog snapshot

Public route, availability, and price checked in the catalog snapshot: Aug 27, 2026, 12:52 AM UTC.

Official documentationKimi reasoning effort guide
Open source
Official documentationKimi K3 Quickstart (Moonshot)
Open source
Official documentationMoonshot model list (retired IDs)
Open source
Official documentationNexoRouter live model page kimi-k3
Open source
NexoRouter catalogNexoRouter published catalog
Aug 27, 2026, 12:52 AM
Official documentationMoonshot AI platform documentation
Open source

Update policy: route evidence targets a 6 h cadence and official-document monitoring, when enabled in production, at least every 24 h. A change creates an internal proposal; claims and translations require human review and are revalidated within 30 days.

For developers

FAQ

What is kimi-k3 for?

kimi-k3 is published for text generation and conversation. Test it against your product's prompts, language, and quality criteria.

What does kimi-k3 cost?

The current public price is $3.04 / 1M tokens for input or generation and $15.20 / 1M tokens for output. The visible unit is the billing unit for this endpoint.

What is this model called in the API?

Use kimi-k3 exactly as the model ID and call /v1/chat/completions with a NexoRouter API key.

How is usage for this model billed?

NexoRouter records usage in the endpoint's public billing unit. The current published figures are $3.04 / 1M tokens for input or generation and $15.20 / 1M tokens for output; the billing page remains the account ledger.

How reliable is this model on NexoRouter?

The current public seven-day aggregate covers 135 attempts with a 98.5% gateway success rate. This is a recent NexoRouter observation, not an upstream SLA.

Why are benchmarks, context, or uptime not shown?

We do not publish figures without a verifiable source, method, and measurement window. When NexoRouter has aggregated telemetry and documentation for those fields, they will be added with their provenance.