⚡ Compare Kimi token rates and calculate your estimated API cost
PAY-AS-YOU-GO · PER 1M TOKENS

Kimi API Pricing

Compare Kimi model costs for input, cached-input and output tokens. Use the interactive calculator to estimate the cost of coding agents, research tools, chatbots and production AI workflows.

Calculate API Cost →Compare Models
$0.10K2.5 cached input
$0.60K2.5 input
$3.00K2.5 output
1MPricing unit
K3 · API pricing and access

Kimi K3 API pricing and Kimi K3 API key

This section covers Kimi K3 API cost, Kimi K3 pricing per token, Kimi K3 token pricing, cost per million tokens, input, output and cached-input prices, API-key setup and possible free-tier access.

Official K3 token rates not yet independently confirmed

Kimi K3 token pricing

Use the live Kimi Developer Platform console as the final source for current K3 availability and billing. The public official sources reviewed for this update did not expose a verified Kimi K3 per-token price table, so this page does not invent rates.

TBCKimi K3 input token price
TBCKimi K3 cached input price
TBCKimi K3 output token price

Kimi K3 cost per million tokens

Once official rates appear, the cost is calculated separately for each one million standard-input, cached-input and output tokens. Smaller usage is billed proportionally.

Kimi K3 pricing per token

Per-token price = published price per one million tokens ÷ 1,000,000. For example, a hypothetical $1.00 per 1M rate equals $0.000001 per token; this is only a formula example, not a K3 rate.

Kimi K3 API free tier

Trial credits, promotional access and regional offers may change. Check the developer console after signing in; do not assume a permanent free API tier unless it is shown on your account.

Pricing accuracy: Kimi Chat, Kimi Code membership and the token-based Developer Platform can use separate billing systems. Confirm the endpoint, model ID, credits and rates shown in your own console before production use.
Get access

How to get a Kimi K3 API key

Open the developer console

Sign in to the official Kimi Developer Platform and open the API-key management area.

Create a new secret key

Generate a key for your project and copy it once. Never publish the secret in HTML, GitHub or client-side JavaScript.

Confirm K3 availability

Check that Kimi K3 appears in your model list and note its exact model ID, endpoint, rate limits and token prices.

Store the key securely

Save it as a server-side environment variable and rotate it immediately if exposed.

export KIMI_API_KEY="your-secret-key"

from openai import OpenAI
client = OpenAI(
  api_key=os.environ["KIMI_API_KEY"],
  base_url="CHECK_OFFICIAL_KIMI_ENDPOINT"
)

# Use the exact K3 model ID shown in your console.
Get Kimi K3 API Key →
K2.7 · API pricing

Kimi K2.7 API pricing

Compare Kimi K2.7 Code API token costs for standard and HighSpeed serving, including input, cached-input and output pricing per one million tokens.

Kimi K2.7 Code · reference pricing

Kimi K2.7 Code API cost

Built for repository analysis, code generation, debugging and agentic software-development workflows. The rates below match the pricing references already used across this page; confirm the live amount in your Developer Platform billing console before deployment.

$0.95Input / 1M tokens
$0.19Cached input / 1M
$4.00Output / 1M tokens

Standard K2.7 Code

Optimized for cost-efficient coding agents and normal production workloads where maximum generation speed is not required.

K2.7 Code HighSpeed

Reference rates: $1.90 input, $0.19 cached input and $8.00 output per 1M tokens for lower-latency, high-throughput coding workflows.

Pricing reminder: API prices, model IDs, availability, rate limits and promotional credits can change. Verify the current Kimi K2.7 API pricing shown in your account before publishing budgets.
Cost breakdown

How Kimi K2.7 token billing works

Kimi K2.7 input token price

Input tokens cover prompts, source files, repository context, tool results and prior conversation content sent to the model.

Kimi K2.7 cached-input price

Eligible repeated context may be billed at the lower cached-input rate, which can reduce costs for reusable system prompts and large codebase context.

Kimi K2.7 output token price

Output tokens include generated code, explanations, reasoning, patches and structured responses returned by the model.

Kimi K2.7 API key

Create a secret key in the Kimi Developer Platform, keep it server-side and use the exact K2.7 model ID and endpoint displayed in your console.

Estimated cost = (input ÷ 1,000,000 × $0.95) + (cached input ÷ 1,000,000 × $0.19) + (output ÷ 1,000,000 × $4.00)
Estimate Kimi K2.7 API cost →
K2.6 · API pricing

Kimi K2.6 API pricing

Explore Kimi K2.6 API cost, token billing, API-key access and pricing considerations for long-horizon coding, multimodal analysis and agent-swarm workflows.

Kimi K2.6 · current model access

Kimi K2.6 token pricing

Kimi K2.6 launched on April 21, 2026 and is available through Kimi products and the Kimi Open Platform. A model-specific public K2.6 per-token price table was not independently verified for this update, so the page avoids presenting an unconfirmed rate.

TBCInput / 1M tokens
TBCCached input / 1M
TBCOutput / 1M tokens
Check live pricing: Sign in to the Kimi Developer Platform and confirm the exact K2.6 model ID, endpoint, input rate, cached-input rate, output rate, credits and regional availability shown for your account.
Billing and access

How Kimi K2.6 API cost works

Kimi K2.6 input token price

Input tokens include prompts, source code, documents, image-derived tokens, system instructions, tool results and conversation history sent to the model.

Kimi K2.6 cached-input price

When prompt caching is supported and repeated context qualifies, cached tokens may be billed at a lower rate than new input. Confirm cache rules in the live platform documentation.

Kimi K2.6 output token price

Output tokens cover generated code, explanations, reasoning, tool calls and structured responses returned by K2.6.

Kimi K2.6 API key

Create a secret key in the official Developer Platform, store it server-side and use the exact K2.6 model name and endpoint shown in your console.

Estimated cost = (input tokens ÷ 1,000,000 × input rate) + (cached tokens ÷ 1,000,000 × cached rate) + (output tokens ÷ 1,000,000 × output rate)
Check Kimi K2.6 API pricing →
01 · Model rates

Kimi API price per 1M tokens

Token billing is based on how much text or multimodal context your application sends and how many tokens the model generates. Cached prompts can materially reduce repeated-context costs.

Kimi K2.5

Multimodal, agent-ready flagship model for vision, coding and complex tool workflows.

BEST VALUE
$0.60Input
$0.10Cached input
$3.00Output

Official K2.5 launch pricing · USD per 1 million tokens.

K2 Thinking Turbo

High-speed reasoning model for long, multi-step and agentic tasks.

REASONING
$1.15Input
$0.15Cached input
$8.00Output

Published Turbo pricing; verify availability and current model IDs in the console.

Kimi K2.7 Code

Coding-focused model for repositories, tool use and software-engineering agents.

CODE
$0.95Input
$0.19Cached input
$4.00Output

Reference rate from the supplied site content; confirm in your billing dashboard.

K2.7 Code HighSpeed

Higher-throughput coding endpoint designed for latency-sensitive agent loops.

HIGHSPEED
$1.90Input
$0.19Cached input
$8.00Output

Reference rate from the supplied site content; confirm before production use.

ModelInput / 1MCached input / 1MOutput / 1MBest for
Kimi K2.5$0.60$0.10$3.00Multimodal agents, coding, vision
K2 Thinking Turbo$1.15$0.15$8.00Fast reasoning and long tasks
Kimi K2.7 Code$0.95$0.19$4.00Software-development agents
K2.7 Code HighSpeed$1.90$0.19$8.00Low-latency coding workflows
Pricing note: Kimi’s model catalogue and rates may change. K2.5 and K2 Thinking Turbo figures are based on published Moonshot/Kimi materials. K2.7 figures are retained from the supplied webpage and should be checked against the live Developer Platform console before publishing or budgeting.
Enterprise · Custom API plans

Kimi enterprise API pricing

Kimi enterprise API pricing is generally suited to organizations with high token volumes, production workloads, security requirements or dedicated support needs. Public fixed enterprise rates were not verified, so this section explains the typical custom-quote structure without inventing prices.

Custom enterprise quote

Enterprise pricing can depend on model choice, monthly token commitment, traffic pattern, context length, throughput, region, support level and deployment requirements.

CONTACT SALES
CustomInput rate
CustomCached input
CustomOutput rate

Ask Moonshot/Kimi for a written quote covering token rates, minimum commitments, overage pricing, taxes and contract term.

What may affect the price?

Large committed usage can sometimes support negotiated economics, but discounts, credits and service terms are account-specific and should not be assumed.

VOLUME
  • Expected monthly input, cached-input and output tokens
  • Standard versus high-speed model serving
  • Concurrency, rate limits and throughput guarantees
  • Long-context, multimodal and agent-tool workloads
  • Support, onboarding and solution-engineering requirements
  • Contract length, billing currency and regional availability

Volume-based API pricing

Prepare realistic monthly token forecasts and peak-request estimates. Ask whether committed usage, prepaid credits or annual contracts change the effective per-million-token rate.

Enterprise SLA and reliability

Request written details for uptime targets, support response times, incident communication, maintenance windows, rate-limit handling and service-credit terms.

Security and governance

Confirm data retention, training-use policy, encryption, access controls, audit logs, key rotation, regional processing, compliance documentation and contractual privacy terms.

Dedicated capacity

For latency-sensitive or high-concurrency workloads, ask about reserved throughput, dedicated capacity, burst limits and whether premium serving has separate token rates.

Enterprise support

Clarify whether the plan includes technical account management, architecture reviews, migration assistance, priority troubleshooting and production launch support.

Billing separation

Kimi Developer Platform API billing is separate from Kimi Chat and Kimi Code memberships. Confirm which console, API key, endpoint and invoice apply to the enterprise deployment.

Request a useful quote
Estimated enterprise spend = committed input usage + cached-input usage + output usage + premium throughput or support charges − negotiated discounts or credits

Send sales a workload summary including model, monthly token volume, peak requests per minute, concurrency, average context size, regions, security requirements and launch date. This makes quotes easier to compare.

Contact Kimi for enterprise pricing →
No public fixed enterprise price verified: Treat any enterprise discount, free credit, dedicated-capacity fee or SLA as valid only when it appears in an official written quote or your Kimi Developer Platform account. Standard API usage is pay-as-you-go and token-based, while enterprise terms may be customized.
02 · Token pricing

Kimi API token pricing explained

Kimi API pricing is measured per one million tokens. Input tokens are the content sent to the model, cached-input tokens are reusable prompt tokens served from cache, and output tokens are generated by the model.

Standard input
$0.60
per 1M Kimi K2.5 tokens

Used for new prompts, uploaded text, code, system instructions and other context processed for the first time.

Cached input
$0.10
per 1M Kimi K2.5 tokens

Applies when eligible repeated context is retrieved from cache, reducing the cost of long reusable prompts and agent instructions.

Model output
$3.00
per 1M Kimi K2.5 tokens

Covers text, code, reasoning, tool-call content and structured responses generated by the model.

Token volumeInput costCached-input costOutput cost
1,000 tokens$0.0006$0.0001$0.0030
10,000 tokens$0.0060$0.0010$0.0300
100,000 tokens$0.0600$0.0100$0.3000
1,000,000 tokens$0.60$0.10$3.00

What counts as a token?

A token is a small unit of text processed by the model. One token may be a word, part of a word, punctuation mark or code fragment.

Why output costs more

Generating responses requires additional inference work, so output-token rates are typically higher than input-token rates.

How to reduce token cost

Shorten prompts, reuse cacheable context, limit unnecessary output and route simpler tasks to the lowest-cost suitable model.

Token-pricing note: The figures above use Kimi K2.5 pricing for simple examples. Other Kimi models have different rates, and current pricing should always be confirmed in the live developer console.
03 · Cost per million tokens

Kimi API cost per million tokens

Kimi API rates are quoted in US dollars for every one million tokens processed. Compare standard input, cached input and generated output costs across the main Kimi models.

Kimi K2.5

Cost-efficient multimodal and general agent workloads.

Input / 1M$0.60
Cached / 1M$0.10
Output / 1M$3.00
K2 Thinking Turbo

Reasoning-intensive and longer multi-step tasks.

Input / 1M$1.15
Cached / 1M$0.15
Output / 1M$8.00
Kimi K2.7 Code

Software development, code analysis and coding agents.

Input / 1M$0.95
Cached / 1M$0.19
Output / 1M$4.00
K2.7 Code HighSpeed

High-throughput, latency-sensitive coding workflows.

Input / 1M$1.90
Cached / 1M$0.19
Output / 1M$8.00

What “per million tokens” means

A rate of $0.60 per one million input tokens means 100,000 input tokens cost $0.06, while 10 million input tokens cost $6.00. Each token category is billed separately.

Example with Kimi K2.5

One million standard input tokens plus one million output tokens costs $3.60. Replacing the input with one million eligible cached tokens reduces the combined cost to $3.10.

Rate reminder: Prices can change and may vary by endpoint, promotion or account. Confirm current Kimi API rates in the official developer console before publishing a budget.
04 · Kimi API cost

How much does the Kimi API cost?

Your Kimi API bill is calculated from three token categories: standard input, cached input and model output. The total changes with the model you choose and the volume of tokens processed.

Cost formula
Total cost = (input tokens ÷ 1,000,000 × input rate) + (cached tokens ÷ 1,000,000 × cached rate) + (output tokens ÷ 1,000,000 × output rate)

For example, using Kimi K2.5 with 10 million input tokens, 5 million cached-input tokens and 2 million output tokens costs $12.50 before taxes, credits or account-specific discounts.

IN

Input token cost

Charged for prompts, documents, code, images converted into model tokens and other context sent to the API.

CA

Cached-input cost

Repeated prompt content may qualify for the lower cached-input rate, helping reduce the cost of reusable system prompts and long context.

OUT

Output token cost

Charged for text, code, reasoning and structured responses generated by the selected Kimi model.

Example monthly costs · Kimi K2.5

Small chatbot

$1.90

1M input + 1M cached input + 0.4M output = $0.60 + $0.10 + $1.20.

Growing AI app

$12.50

10M input + 5M cached input + 2M output = $6.00 + $0.50 + $6.00.

Production workflow

$77.00

50M input + 20M cached input + 15M output = $30.00 + $2.00 + $45.00.

Cost-saving tip: Keep prompts concise, reuse cacheable context, select the lowest-cost model that meets quality requirements and monitor output-token limits.
Calculate your Kimi API cost →
05 · Cost estimator

Kimi API pricing calculator

Enter the estimated monthly token volume. The calculator divides each amount by one million and applies the selected model’s rate.

Estimated monthly cost
$12.50

This estimate excludes taxes, credits, minimum charges and any account-specific discounts.

Input cost$6.00
Cached-input cost$0.50
Output cost$6.00
06 · Quick start

Start using the Kimi API

01

Create an account

Open the Kimi Developer Platform and create or sign in to your developer account.

02

Add billing

Fund the API account or configure its supported billing method. API billing is separate from Kimi Chat and Kimi Code memberships.

03

Generate a key

Create an API key, store it securely and send requests through an OpenAI-compatible client.

Python · OpenAI-compatible example
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_KIMI_API_KEY",
    base_url="https://api.moonshot.ai/v1"
)

response = client.chat.completions.create(
    model="kimi-k2.5",
    messages=[
        {"role": "user", "content": "Explain cached-input pricing."}
    ]
)

print(response.choices[0].message.content)
API pricing comparison

Kimi API pricing comparison table

Compare representative Kimi, OpenAI, Claude, Gemini, DeepSeek and Grok API token prices. Rates are shown in USD per one million tokens and may vary by context length, batch mode, region, endpoint or account.

Checked July 18, 2026
ProviderRepresentative modelInput / 1MCached input / 1MOutput / 1MPricing note
KimiKimi K3TBCTBCTBCOfficial public K3 token table not verified; check the Kimi console.
KimiKimi K2.5$0.60$0.10$3.00Published launch pricing used as the verified Kimi reference.
OpenAIGPT-5.4$2.50$0.25$15.00Long prompts above the listed threshold can use higher rates.
AnthropicClaude Sonnet 4.6$3.00$0.30$15.00Cache-read rate shown; cache writes are priced separately.
GoogleGemini 3.1 Pro Preview$2.00$0.20$12.00Standard rate for prompts up to 200K tokens.
DeepSeekDeepSeek V4 Pro$0.435$0.003625$0.87Cache-miss input, cache-hit input and output rates.
xAIGrok 4.5$2.00$0.50$6.00Short-context pricing; long-context requests cost more.

Kimi API vs OpenAI API pricing

Using Kimi K2.5 as the verified Kimi reference, its $0.60 input and $3 output rates are lower than GPT-5.4 at $2.50 input and $15 output. Model capability and workload quality should be evaluated alongside token price.

Kimi API vs Claude API pricing

Kimi K2.5 is cheaper per listed token than Claude Sonnet 4.6. Claude uses separate cache-write and cache-read pricing, so repeated-context workloads need a full caching calculation.

Kimi API vs Gemini API pricing

Kimi K2.5 has lower listed standard input and output rates than Gemini 3.1 Pro Preview. Gemini also offers free-tier access for selected models and discounted batch or flex processing.

Kimi API vs DeepSeek API pricing

DeepSeek V4 Pro has the lowest rates in this representative table, especially for cache hits. Compare model quality, availability, latency, context limits and data requirements before choosing only by price.

Kimi API vs Grok API pricing

Kimi K2.5 is lower-priced than Grok 4.5 for standard input, cached input and output. Grok 4.5 also applies higher long-context rates once its threshold is exceeded.

Kimi K3 vs GPT, Claude and Gemini API cost

A factual Kimi K3 cost comparison is not yet possible without verified K3 input, cached-input and output rates. Once Moonshot publishes them, compare the same token category, context tier and processing mode.

Comparison basis: OpenAI GPT-5.4, Anthropic Claude Sonnet 4.6, Google Gemini 3.1 Pro Preview, DeepSeek V4 Pro and xAI Grok 4.5 official pricing pages. Kimi K3 remains marked TBC because a verified public K3 token table was not available. Prices can change; always confirm live billing before deployment.
Important: This table compares representative models, not identical capability tiers. Tool calls, search grounding, images, audio, cache storage, batch processing, long context and taxes can add or change costs.
07 · FAQ

Kimi API pricing questions

How much does the Kimi API cost?

Cost depends on the model and the number of input, cached-input and output tokens. Kimi K2.5 launch pricing is $0.60, $0.10 and $3.00 respectively per one million tokens.

What is cached-input pricing?

When identical or reusable prompt content is served from cache, those tokens can be charged at a lower cached-input rate instead of the full input rate.

Is there a free Kimi API tier?

Promotional credits or trial access may vary by account and region. Check the live console for current offers rather than assuming ongoing free API usage.

Is Kimi membership included with API billing?

No. Kimi Chat and Kimi Code subscriptions are separate from the Developer Platform’s pay-as-you-go, token-based API system.

Which model is cheapest?

Among the listed models, Kimi K2.5 has the lowest published input, cached-input and output rates, making it a strong general-purpose value option.

Can I use the OpenAI SDK?

Yes. Configure the supported Kimi base URL, API key and model name in an OpenAI-compatible SDK. Always confirm the current endpoint and model ID in official documentation.