Kimi API Pricing 2026: Plans & Cost Guide

The Moonshot Kimi API is a pay-as-you-go large language model API from Moonshot AI, priced per million tokens across five models with context windows ranging from 8K to 128K tokens. The table below shows current pricing for all models in both Chinese Yuan (CNY) and approximate USD.

USD prices are approximate conversions for reference. Official Kimi API pricing is denominated in Chinese Yuan (CNY) at platform.moonshot.cn. Actual USD charges depend on your payment processor's exchange rate.

⚠️ Prices verified as of June 2025. Kimi API pricing changes frequently. Always confirm current rates at the Moonshot AI API pricing page at platform.moonshot.cn/pricing before budgeting.

Last Updated: June 2025

On this page:

• What Is the Moonshot Kimi API?
• Kimi API Models and Pricing (2026)
• Kimi API Free Tier: What You Get for Free
• Rate Limits by Tier
• How to Calculate Your Kimi API Costs
• Kimi API vs. Competitors: Pricing Comparison
• How to Reduce Your Kimi API Costs
• Enterprise and Volume Pricing
• Is the Kimi API Available Outside China?
• How to Get Started with the Kimi API
• Frequently Asked Questions

What Is the Moonshot Kimi API?

Moonshot AI (月之暗面) is a Beijing-based AI startup founded in 2023, reportedly raising over $1 billion in funding by 2024, and positioning itself as a Chinese competitor to OpenAI and Google in the large language model space. The company's developer product is the Kimi API at platform.moonshot.cn, which is a separate product from the consumer chatbot at kimi.ai. This guide covers the developer API only.

The Kimi API uses an OpenAI-compatible REST format, which means most OpenAI SDK examples require only minimal changes to work with Kimi. That makes it a practical migration path for developers already running on GPT-4o.

Token: The basic unit of text billed by LLM APIs. In English, 1 token ≈ 0.75 words, so 1,000 words ≈ 1,333 tokens. Both the tokens you send to the model (input tokens) and the tokens in the model's response (output tokens) are metered and billed separately.

The context window is the maximum number of tokens a model can process in a single API call, covering both the input and the output combined. The moonshot-v1-128k model supports 128K tokens, which equals approximately 96,000 words or the length of a full novel, in one call.

Kimi uses a pay-as-you-go billing model with no mandatory monthly subscription. You pay only for the tokens you consume, and there are no fixed platform fees.

MOONSHOT token on Bybit: Moonshot AI's expansion into frontier AI has attracted attention from crypto traders — Bybit offers MOONSHOTUSDT perpetual futures for those interested in trading the MOONSHOT token. See also Moonshot AI's IPO trading opportunities on Bybit.

Kimi API Models and Pricing (2026)

Moonshot offers five models via the Kimi API: three standard chat models differentiated by context window size, one reasoning model, and one vision-language model.

💡 Total Cost = (Input Tokens × Input Price per MTok) + (Output Tokens × Output Price per MTok)

Input tokens are the tokens in the message you send to the model: your system prompt, user message, and any conversation history. Output tokens are the tokens in the model's generated response. Kimi prices input and output tokens identically across the moonshot-v1 series, which differs from OpenAI and Anthropic where output tokens cost 3–5× more than input tokens. This symmetry makes Kimi billing easier to estimate for applications that generate long responses.

For chatbots, input token counts grow with conversation history each turn. For document summarization, output costs are lower relative to the large input you send. Plan your token budget accordingly.

Streaming responses do not affect pricing. Billed tokens are identical whether streaming is enabled or disabled.

ℹ️ All CNY prices are official Moonshot pricing. USD prices shown in tables are approximate conversions. Official pricing is denominated in CNY at platform.moonshot.cn. Divide any per-MTok price by 1,000 to get the per-1K-token rate.

The Kimi API uses pay-as-you-go pricing with no monthly subscription plans. There are no fixed tier fees. You pay for what you use.

moonshot-v1-8k

moonshot-v1-8k is the lowest-cost Kimi model, with an 8,192-token context window suited to short Q&A, classification tasks, and single-turn chatbot responses where long context is not needed.

The "8k" suffix denotes the context window size in tokens, not a version or quality tier. Use this model when your prompts and responses fit comfortably within 8,192 tokens and you want the lowest per-token rate.

Best for: Single-turn Q&A, intent classification, short text generation, customer support bots with brief exchanges.

moonshot-v1-32k

moonshot-v1-32k offers a 32,768-token context window, which is sufficient for multi-turn conversations, medium-length document review, and most code review tasks without paying the full premium of the 128K model.

Choose this model when your use case exceeds 8K context but doesn't require processing full long-form documents in a single call.

Best for: Multi-turn chat sessions, code review, medium-length documents, email drafting with context.

moonshot-v1-128k

moonshot-v1-128k supports a 131,072-token context window, which equals approximately 96,000 words or the length of a full novel. It is Kimi's primary choice for full-document analysis, legal contract review, and large codebase processing. At ¥60 CNY (~$8.25 USD) per MTok, this model is significantly cheaper than GPT-4o at the same 128K context window size.

This is the model most developers cite when comparing Kimi against OpenAI for long-context workloads. The 128K capacity covers approximately 300 pages of technical documentation in a single call.

Best for: Full document analysis, legal contract review, research paper summarization, large codebase processing, book-length text analysis.

Kimi k1.5 (Reasoning Model)

Kimi k1.5 is a separate reasoning-focused model, not a larger version of the moonshot-v1 series. It is designed for chain-of-thought tasks including complex mathematics, multi-step coding challenges, and research analysis. Reasoning models like Kimi k1.5 compete directly with OpenAI o1/o3 and DeepSeek-R1 in the extended thinking category.

Pricing for Kimi k1.5 should be verified at platform.moonshot.cn/docs before budgeting, as reasoning models often carry higher per-token rates than standard chat models due to the additional compute used for internal chain-of-thought processing. Some reasoning models also charge separately for "thinking tokens." Confirm whether this applies to Kimi k1.5 in the current documentation.

Best for: Complex multi-step reasoning, advanced mathematics, difficult coding challenges, research synthesis requiring logical chains.

For a full technical review of Kimi K3's capabilities and benchmarks, see Kimi K3: Moonshot AI's Flagship Reasoning Model.

Kimi VL (Vision-Language Model)

Kimi VL accepts both image and text input in a single API call, with text tokens priced per MTok and image inputs priced separately based on resolution. This is Moonshot's multimodal vision model for tasks that require reading images alongside text.

Image pricing for Kimi VL typically follows a per-tile or per-image methodology where higher-resolution images consume more tokens. Verify the current image pricing methodology at platform.moonshot.cn/docs before building vision-based applications, as image token rates differ from text token rates.

Best for: Document OCR, image analysis, visual Q&A, diagram interpretation, screenshot-to-text extraction.

⚠️ Prices verified as of June 2025. All prices subject to change. Confirm at platform.moonshot.cn before budgeting.

Kimi API Free Tier: What You Get for Free

✅ Free Tier Summary: New accounts receive approximately ¥15 CNY in free API credits upon registration. Credits expire after a limited period (verify the current expiration date at platform.moonshot.cn before signing up, as terms change). Free tier rate limits apply. All moonshot-v1 models are accessible with free credits.

Moonshot allocates free trial credits to new accounts upon registration. This is a one-time credit allocation, not a time-limited free trial in the SaaS sense. You are getting a fixed credit balance to spend on API calls, not 30 days of unlimited access.

At ¥12 CNY per MTok on moonshot-v1-8k, ¥15 CNY in free credits covers approximately 1.25 million tokens of input or output. That is enough to run several hundred test requests and prototype a basic integration before committing any payment details.

Free tier rate limits are suitable for testing and prototyping but not for production traffic. If you need higher throughput, you will need to add credits and move to a paid tier. The free tier applies to all three moonshot-v1 models, so you can test the 8K, 32K, and 128K variants with your free credits to compare output quality before choosing a model.

To claim your free credits:

1. Register at platform.moonshot.cn
2. Verify your email or phone number
3. Free credits are credited to your account automatically upon verification

Claim your free Kimi API credits at platform.moonshot.cn.

Rate Limits by Tier

Rate limits control how fast you can call the Kimi API, measured in RPM (Requests Per Minute) and TPM (Tokens Per Minute). They govern call speed, not total monthly volume.

Verify current rate limit values at platform.moonshot.cn/docs before building production applications. Rate limit tiers and thresholds change as Moonshot updates its platform.

Rate limits typically increase as your cumulative account spend grows. Moonshot ties higher limits to spend history. Accounts that have spent more on the platform receive progressively higher RPM and TPM allocations without requiring a separate plan upgrade.

⚠️ Exceeding rate limits returns an HTTP 429 Too Many Requests error. Implement exponential backoff retry logic in your application to handle rate limit responses gracefully. Start with a 1-second delay and double on each subsequent retry up to a maximum wait time.

To request a quota increase beyond your current tier, contact Moonshot support via the platform dashboard at platform.moonshot.cn or through the enterprise inquiry pathway described in the Enterprise and Volume Pricing section below.

Rate limits are distinct from spending limits. Rate limits govern how fast you can make calls, not how much total credit you can spend. Pre-loading a specific credit balance is the primary mechanism for controlling total spend.

How to Calculate Your Kimi API Costs

💡 Total Cost = (Input Tokens × Input Price per MTok) + (Output Tokens × Output Price per MTok)

Input tokens are the tokens in the message you send to the model. Output tokens are the tokens in the model's response. Both are metered separately and billed at the rates shown in the pricing table above. On the moonshot-v1 series, input and output are priced identically, so the formula simplifies to: Total Tokens × Price per MTok.

Word-to-token conversion reference:

To calculate your monthly cost, follow these steps:

1. Estimate your average input tokens per request (system prompt + user message + history).
2. Estimate your average output tokens per request (model response length).
3. Add input and output tokens to get total tokens per request.
4. Multiply by your requests per day to get daily token volume.
5. Apply the per-MTok rate for your chosen model: Total Tokens ÷ 1,000,000 × Price per MTok.
6. Multiply by 30 for a monthly estimate.

Use this table to estimate your input size from word counts. For more precise pre-call token counting, use OpenAI tiktoken (github.com/openai/tiktoken) as an approximation. Kimi's tokenizer may differ slightly, so check platform.moonshot.cn/docs for Moonshot's official tokenizer endpoint.

Worked Example A: Chatbot (Medium Volume)

Scenario: Customer support chatbot handling 1,000 queries/day, each with 800 input tokens and 400 output tokens. Model: moonshot-v1-8k.

• Daily input tokens: 1,000 × 800 = 800,000 tokens = 0.8 MTok
• Daily output tokens: 1,000 × 400 = 400,000 tokens = 0.4 MTok
• Daily total: 1.2 MTok × ¥12 CNY = ¥14.40 CNY/day (~$1.98 USD/day)
• Monthly cost: ~¥432 CNY (~$59 USD)

Worked Example B: Document Summarizer (Low Volume)

Scenario: Research tool processing 100 documents/day, each with 10,000 input tokens and 500 output tokens. Model: moonshot-v1-128k (required for long documents).

• Daily input tokens: 100 × 10,000 = 1,000,000 tokens = 1 MTok
• Daily output tokens: 100 × 500 = 50,000 tokens = 0.05 MTok
• Daily total: 1.05 MTok × ¥60 CNY = ¥63 CNY/day (~$8.67 USD/day)
• Monthly cost: ~¥1,890 CNY (~$260 USD)

Worked Example C: Code Assistant (Medium Volume)

Scenario: Developer tool making 500 requests/day, each with 2,000 input tokens and 800 output tokens. Model: moonshot-v1-32k.

• Daily input tokens: 500 × 2,000 = 1,000,000 tokens = 1 MTok
• Daily output tokens: 500 × 800 = 400,000 tokens = 0.4 MTok
• Daily total: 1.4 MTok × ¥24 CNY = ¥33.60 CNY/day (~$4.62 USD/day)
• Monthly cost: ~¥1,008 CNY (~$139 USD)

Kimi's billing operates on a credit prepay model. You load credits to your account balance and API calls draw down that balance. For ongoing cost tracking, use Artificial Analysis (artificialanalysis.ai) to benchmark Kimi against other providers with up-to-date USD pricing.

Kimi API vs. Competitors: Pricing Comparison

The table below compares Kimi API pricing against OpenAI, DeepSeek, Anthropic, and Google Gemini, all in USD per million tokens as of June 2025.

⚠️ All prices verified as of June 2025. LLM provider pricing changes frequently. Verify all prices at official documentation pages before making vendor decisions.

The Kimi moonshot-v1-128k model costs approximately $8.25 per MTok for both input and output, compared to GPT-4o at $2.50 input / $10.00 output per MTok for the same 128K context window. For applications generating substantial output, Kimi is materially cheaper.

When Kimi wins: Long-context workloads at 128K where GPT-4o output costs become significant, or applications with symmetric input/output volumes where Kimi's identical input/output pricing simplifies budgeting.

When competitors may win: DeepSeek-V3 undercuts Kimi significantly on per-token rates ($0.27 input vs. ~$1.65 for Kimi 8K) for short-context tasks. Google Gemini 1.5 Pro offers a 1M-token context window at $1.25/$5.00 per MTok, which is a better fit for workloads exceeding 128K tokens. Anthropic Claude 3.5 Sonnet provides a 200K context window with strong reasoning capability for teams that need the extra context headroom beyond Kimi's 128K.

Chinese developers comparing within the domestic LLM ecosystem may also evaluate Qwen from Alibaba Cloud, which offers competitive pricing for Chinese-language tasks.

For a broader model-to-model comparison including benchmarks and use-case fit, see Moonshot AI vs DeepSeek vs Qwen: China AI Models Compared.

How to Reduce Your Kimi API Costs

The single largest cost lever in the Kimi API is model selection: using moonshot-v1-8k instead of moonshot-v1-128k for tasks that fit within an 8K context window reduces per-token spend by 5×. Review the following strategies before scaling any Kimi-powered application.

1. Choose the right model tier. Audit your actual prompt and response lengths. If your average request uses under 6,000 tokens, moonshot-v1-8k covers it at ¥12 CNY/MTok rather than ¥60 CNY/MTok. The price difference compounds at scale.

2. Truncate conversation history. In multi-turn chatbot applications, re-sending the full conversation history each turn accumulates large input token counts. Keep only the last 3–5 exchanges in the context window instead of the full session transcript.

3. Compress your prompts. Remove redundant instructions, repeated context, and verbose phrasing from system prompts. A 500-token system prompt that can be reduced to 200 tokens saves 300 tokens per request. At 100,000 requests/month, that's 30 million tokens in savings.

4. Cache repeat responses. If your application sends identical or near-identical queries frequently (FAQ bots, product description generators with the same base prompt), cache the response and skip the API call entirely for repeated inputs.

5. Set a spending cap. Pre-load a specific credit amount to your Kimi account balance rather than enabling automatic top-up. This acts as a hard spending cap. When the balance reaches zero, API calls stop. Check your balance regularly via the platform.moonshot.cn dashboard.

6. Use the max_tokens parameter. Set max_tokens in your API request to limit output length for tasks that don't require long responses. Classification tasks, yes/no answers, and extraction tasks rarely need more than 100–200 output tokens.

7. Monitor usage in the billing dashboard. Review token consumption by model and date in the platform.moonshot.cn dashboard. Identifying which endpoints consume the most tokens often reveals optimization opportunities that aren't apparent from code inspection alone.

Enterprise and Volume Pricing

Moonshot AI does not publicly list enterprise or volume discount pricing. The rates in this article reflect publicly available pay-as-you-go pricing only. For enterprise contracts, dedicated rate limits, or volume discounts, contact Moonshot AI directly through the enterprise inquiry form at platform.moonshot.cn.

Organizations evaluating Kimi at scale typically need to assess: dedicated rate limit ceilings above the standard paid tiers, volume discount thresholds tied to monthly token spend, data residency guarantees and compliance certifications, SLA terms and uptime commitments, and contractual data handling agreements.

None of these terms are publicly documented at time of writing. Do not assume volume discounts exist at any specific spend threshold. Verify directly with Moonshot's sales team before building financial projections around discount assumptions.

Is the Kimi API Available Outside China?

Yes, the Kimi API is accessible to developers outside China via platform.moonshot.cn. International developers can create accounts and generate API keys using the standard registration flow, which accepts international email addresses.

Payment method acceptance for non-Chinese users should be verified at registration time. Accepted methods may include international credit cards or third-party payment processors. Confirm supported payment options at platform.moonshot.cn before attempting to add billing credentials.

The sign-up process for international users follows the same steps as for Chinese users: create an account, verify your phone number or email, and generate an API key from the dashboard. Some international phone numbers may not be accepted for SMS verification. If you encounter this, use email verification or contact Moonshot support.

🔒 Data Privacy Notice: Moonshot AI is a Chinese company and the Kimi API infrastructure operates under Chinese law and data regulations. API request data is processed on Moonshot AI's servers in China. Developers in regulated industries or jurisdictions with data residency requirements should review Moonshot's privacy policy at platform.moonshot.cn before deploying the Kimi API in production environments. Do not send data subject to GDPR, HIPAA, or other regulated data frameworks without reviewing Moonshot's compliance documentation and data processing agreements.

For enterprise users with compliance requirements, the absence of publicly documented data residency options, SOC 2 certification, or data processing agreements is a material consideration. Direct any compliance inquiries to Moonshot's enterprise team. This is not a reason to dismiss Kimi for all use cases. Many developer applications handle non-sensitive data where Chinese jurisdiction presents no compliance barrier. The point is deliberate evaluation rather than assumption.

How to Get Started with the Kimi API

Creating a Kimi API account and generating your first API key takes under five minutes using the following steps.

1. Create an account at platform.moonshot.cn. International users may need to provide a phone number for SMS verification. Have an international mobile number ready, or use email-based registration if phone verification is unavailable in your country.

2. Verify your email or phone number. Complete the verification step to activate your account and unlock free trial credits.

3. Claim your free trial credits. New accounts receive approximately ¥15 CNY in free credits automatically upon verification. Check your account balance in the dashboard to confirm the credit has been applied.

4. Navigate to "API Keys" in the dashboard. The API key management section is accessible from the main dashboard menu after login.

5. Generate a new API key and copy it. Click "Create API Key," give it a descriptive name, and copy the key immediately. It will not be shown again in full after the creation screen.

6. Store the API key as an environment variable. Add MOONSHOT_API_KEY=your_key_here to your .env file or your deployment environment's secret manager. Never hardcode the key directly in source files or commit it to version control.

7. Review the API documentation. The Kimi API uses an OpenAI-compatible format. The base URL changes but the request structure is identical to the OpenAI Chat Completions API. Most OpenAI SDK examples work with minimal modification. Access the full documentation at platform.moonshot.cn/docs.

Get your free Kimi API key at platform.moonshot.cn.

Frequently Asked Questions

How much does the Kimi API cost?

The Kimi API costs between ¥12 CNY (~$1.65 USD) per million tokens for the moonshot-v1-8k model and ¥60 CNY (~$8.25 USD) per million tokens for the flagship moonshot-v1-128k model. Input and output tokens are priced identically across the moonshot-v1 series, which is a different structure from OpenAI where output tokens cost 4× more than input. All pricing is pay-as-you-go with no mandatory subscription fee.

Does the Kimi API have a free tier?

Yes. New accounts receive approximately ¥15 CNY in free trial credits upon registration at platform.moonshot.cn. These credits apply to all moonshot-v1 models and allow you to prototype an integration before adding payment credentials. The free credits are a one-time allocation, not an ongoing free tier. Once spent, you need to add paid credits to continue. Free tier rate limits (approximately 3 RPM) are suitable for testing but not for production traffic. See the Kimi API Free Tier section above for full details.

What is the maximum context window for the Kimi API?

The largest publicly available Kimi API model, moonshot-v1-128k, supports a 128K token context window, which equals approximately 96,000 words or a full-length novel. Kimi also offers moonshot-v1-8k (8,192 tokens) and moonshot-v1-32k (32,768 tokens) at lower price points. Moonshot has demonstrated experimental 1M-token context in research, but this is not currently available via the public API.

How does Kimi API pricing compare to OpenAI GPT-4o?

Kimi is cheaper than GPT-4o for comparable context window sizes, particularly for output-heavy workloads. GPT-4o costs $2.50 per MTok for input and $10.00 per MTok for output at 128K context, while Kimi moonshot-v1-128k costs approximately $8.25 per MTok for both input and output. For applications generating substantial output, the effective cost difference is material. For input-heavy workloads with short outputs, GPT-4o mini at $0.15/$0.60 per MTok is cheaper than any Kimi model. See the full pricing comparison table above.

Is the Kimi API available outside China?

Yes. The Kimi API is accessible globally via platform.moonshot.cn. International developers can register, add payment credentials, and use the API outside China through the standard sign-up flow. Data is processed on Moonshot AI's infrastructure in China. Developers in regulated industries with data residency requirements should review Moonshot's privacy policy before deploying in production. See the international availability section above for the full data privacy notice.

What models does Moonshot Kimi offer via API?

Moonshot offers five models via the Kimi API: moonshot-v1-8k (8K context, lowest cost), moonshot-v1-32k (32K context, mid-tier), moonshot-v1-128k (128K context, flagship), Kimi k1.5 (reasoning model with chain-of-thought capability, competing with OpenAI o1), and Kimi VL (vision-language model accepting image and text input). Each model has separate pricing, with the moonshot-v1 series publicly listed and Kimi k1.5/VL pricing requiring verification at platform.moonshot.cn/docs.

How do I calculate my Kimi API costs?

Use the formula: Total Cost = (Input Tokens × Input Price per MTok) + (Output Tokens × Output Price per MTok). For example, a chatbot sending 800 input tokens and receiving 400 output tokens per request on moonshot-v1-8k costs (800 × ¥12/1,000,000) + (400 × ¥12/1,000,000) = ¥0.0144 CNY per request (~$0.002 USD). At 1,000 requests/day, that totals approximately ¥432 CNY (~$59 USD) per month. For full worked examples at different usage scales, see the How to Calculate Your Kimi API Costs section above.

What are the rate limits for the Kimi API?

Rate limits vary by account tier. The free tier is limited to approximately 3 RPM (Requests Per Minute) and 32,000 TPM (Tokens Per Minute), which is sufficient for testing but not production traffic. Standard paid tiers offer 60–300 RPM and 200,000–1,000,000 TPM, scaling with cumulative account spend. Exceeding rate limits returns an HTTP 429 Too Many Requests error. See the Rate Limits by Tier section above for the full tier comparison table.

Is the Kimi API cheaper than GPT-4?

Yes. moonshot-v1-128k costs approximately $8.25 per MTok for both input and output, considerably less than GPT-4o's $10.00 per MTok for output at the equivalent 128K context window. For symmetric input/output workloads, the savings are substantial. However, DeepSeek-V3 offers even lower prices ($0.27/$1.10 per MTok) for short-context tasks, and Google Gemini 1.5 Pro provides a 1M-token context window at $1.25/$5.00 per MTok for applications exceeding 128K context. Kimi's price advantage is strongest for long-context, output-heavy workloads where GPT-4o output costs dominate the bill.

How do I get a Kimi API key?

Create an account at platform.moonshot.cn, verify your email or phone number, navigate to the "API Keys" section of the dashboard, and generate a new key. Copy it immediately from the creation screen. Store the key as an environment variable (MOONSHOT_API_KEY) in your project. Never hardcode it in source files. New accounts also receive free trial credits automatically upon verification.

Kimi API pricing positions moonshot-v1-128k as a cost-competitive option for long-context workloads where GPT-4o or Claude become expensive at scale, particularly for output-heavy applications that benefit from Kimi's identical input/output pricing structure.

View the current Kimi API pricing at platform.moonshot.cn.

Last updated: June 2025 | Prices sourced and verified at platform.moonshot.cn/docs

Related Reading

• Kimi K3: Moonshot AI's Flagship Reasoning Model
• What Is Kimi K2? Moonshot AI's Open-Weight Thinking Model
• Moonshot AI: Company Overview and Kimi Product Guide
• Moonshot AI vs DeepSeek vs Qwen: China AI Models Compared
• What Is Kimi AI? Guide to Moonshot AI
• MOONSHOTUSDT Perpetual Futures on Bybit