Kimi API · Ace Data Cloud

Kimi K3 API
Flagship Model for Programming & Agents

Call Kimi K3 using the OpenAI-compatible format. 1M Token context, native visual understanding, and long-horizon programming capabilities, with reasoning effort controlled via reasoning_effort.

📄 View Documentation 💰 View Pricing

🚀 Kimi K3 🧠 Max Reasoning 🔌 OpenAI Compatible 🌐 Chinese-English Bilingual

🤖

Flagship Model

🧠

2.8T

Total MoE Parameters

💰

Low Price

Lower Than Official Pricing

⚡

Context Window

Why Use Kimi Through Ace Data Cloud?

Kimi K3 is Moonshot AI's flagship model, designed for long-horizon programming, complex reasoning, and end-to-end knowledge work. It features a 1M Token context window, enabling sustained understanding of large codebases and coordination of tools to complete multi-step tasks.

Ace Data Cloud provides Kimi K3 and K2 series APIs in an OpenAI-compatible format. Existing OpenAI SDKs can connect by simply changing the base URL and model name.

Core Capabilities of the Kimi K3 API

A unified interface for complex programming, agents, and knowledge work

🔌

OpenAI-Compatible Format

Call Kimi K3 via /v1/chat/completions, compatible with the OpenAI SDK, streaming responses, and tool calls.

🧠

Max Reasoning Mode

K3 always enables reasoning. The only currently supported value for reasoning_effort is max, suitable for complex programming, mathematics, and agent tasks.

🌐

Chinese-English Bilingual Optimization

Excellent native Chinese understanding, with outstanding English performance as well. Ideal for Chinese-language scenarios, cross-language translation, and multilingual application development.

💻

Top-Tier Programming Capabilities

K3 is designed for long-horizon programming tasks, capable of understanding large codebases, coordinating terminal tools, and completing code generation, debugging, and review.

⚡

Streaming

Supports SSE streaming with real-time token-by-token output. Set stream: true to enable streaming responses.

📄

Structured Output

Specify the JSON output format with response_format, and use JSON Schema to obtain structured API response data.

Python

from openai import OpenAI
client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.acedata.cloud/v1"
)
response = client.chat.completions.create(
  model="kimi-k3",
    messages=[
    {"role": "system", "content": "请简洁回答，只输出经过测试的 Python 代码，不要解释；确保重复元素不会导致无限递归。"},
        {"role": "user", "content": "用 Python 实现一个快速排序算法"}
    ],
  reasoning_effort="max",
  stream=False
)
print(response.choices[0].message.content)

Actual Response (Condensed)

{
  "id": "msg_HvLVj1NEhTi0osEguZllkYpI",
  "object": "chat.completion",
  "created": 1784466676,
  "model": "kimi-k3",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "```python\ndef quicksort(arr):\n    if len(arr) <= 1:\n        return arr\n    pivot = arr[len(arr) // 2]\n    left = [x for x in arr if x < pivot]\n    middle = [x for x in arr if x == pivot]\n    right = [x for x in arr if x > pivot]\n    return quicksort(left) + middle + quicksort(right)\n```"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 128,
    "completion_tokens": 1493,
    "total_tokens": 1621
  }
}

Switch with One Line of OpenAI SDK Code

Simply replace base_url and model to use Kimi K3 in your existing OpenAI project, and adjust reasoning effort through reasoning_effort.

Get an API Key

Modify base_url

Set base_url to https://api.acedata.cloud/v1

Select a Kimi Model

Set model to kimi-k3; to explicitly configure reasoning parameters, use reasoning_effort="max"

What Can You Build with the Kimi API?

From Chinese NLP to long-horizon programming—developers are building these applications with Kimi K3

💬

Chinese Conversational Assistants

Build high-quality Chinese customer service, knowledge Q&A, and personal AI assistants with native Chinese understanding that far surpasses general-purpose models

💻

Code Generation and Review

K3 supports code generation, bug fixes, code review, architecture recommendations, and multi-step terminal tool collaboration

🌐

Cross-Language Translation

Native Chinese-English bilingual capabilities, ideal for high-quality translation, localization, and multilingual content generation

🔬

Deep Reasoning and Analysis

Math problem solving, logical reasoning, data analysis—Thinking mode provides step-by-step reasoning

Get Started Quickly in 3 Steps

From registration to sending your first Kimi K3 message, it takes less than 3 minutes

Register and Get an API Key

Create a free account on Ace Data Cloud and generate your Bearer Token from the console.

Call It Using the OpenAI SDK

Set base_url to Ace Data Cloud and select kimi-k3 to get started.

Integrate and Scale

Embed Kimi K3 in your application. The OpenAI-compatible format makes switching between models effortless.

View Documentation →

Why choose Ace Data Cloud instead of using the Moonshot API directly?

Comprehensive advantages in format compatibility, global availability, and unified interfaces

Comparison dimension	Ace Data Cloud	Direct Moonshot connection
OpenAI-compatible format	✓	Partially compatible
Available globally	✓ Ready to use out of the box	✗ Restricted in some regions
Streaming	✓	✓
Unified interface for multiple models	✓ GPT / Claude / Gemini / Kimi	✗ Kimi only
Pay as you go	✓ Flexible top-ups	✓
No overseas mobile number required for registration	✓	✗ Chinese mobile number required
Thinking reasoning model	✓	✓

Choose the right Kimi model

K3 is suited for the most complex programming and Agent tasks, while K2.6 and K2.5 cover general scenarios

K2.6

Efficient general-purpose

A general-purpose model that supports reasoning, tool calling, and multimodal input, suitable for everyday conversations and Agent tasks.

✓ 256K context
✓ Switchable Thinking
✓ Suitable for general conversations and analysis
✓ Supports tool calling

Recommended

Kimi K3

Flagship model

A flagship model for long-horizon programming and complex Agent workflows, with reasoning always enabled and currently supporting Max reasoning mode.

✓ 1M Token context
✓ reasoning_effort: max
✓ Tool calling and streaming output
✓ Top choice for complex programming tasks

K2.5

Compatible

A stable K2 series model, suitable for general reasoning and content generation requiring 256K context.

✓ 256K context
✓ Switchable Thinking
✓ Enhanced instruction following
✓ More precise output control

kimi-k3 kimi-k2.6 kimi-k2-thinking kimi-k2-thinking-turbo kimi-k2.5

Kimi API Pricing

Pay by Token usage. No subscription fees, no hidden charges.

Volume packages offer greater discounts

Pay as you go

Token Billing

Low Price billed by Token

Charged based on actual Token usage, with separate pricing for input and output

✓ Transparent model-based pricing for K3, K2.6, and K2.5
✓ Reasoning Tokens billed based on actual usage
✓ No subscription fees or minimum spend
✓ Input and output priced separately
✓ Streaming — Free

View Pricing Details View API Documentation

Enterprise

Custom

Dedicated solutions for high-volume teams

✓ Volume-based tiered discounts
✓ Priority support and account manager
✓ Custom rate limits
✓ SLA guarantee
✓ Private deployment solutions

Contact Sales

Frequently Asked Questions

Everything you need to know about using the Kimi K3 API

What tasks is Kimi K3 suitable for? ▾

Kimi K3 is Moonshot AI's flagship MoE model, with 2.8 trillion total parameters and a 1M Token context window, making it suitable for large codebase understanding, long-horizon programming, complex reasoning, tool calling, and end-to-end knowledge work.

Does it support the OpenAI SDK? ▾

Yes. When using the OpenAI SDK, set base_url to https://api.acedata.cloud/v1 and model to kimi-k3. K3 also supports tool calling, streaming responses, and reasoning_effort.

How should reasoning_effort be configured? ▾

K3 always enables reasoning, and the only currently supported value is max; omitting it also uses max. Some upstream-compatible implementations may accept other strings, but they are not guaranteed to change reasoning behavior, so they are not recommended.

Official references: Thinking Effort, Model Parameter Reference, and Chat API.

How do its Chinese capabilities compare with GPT/Claude? ▾

The Kimi series has native advantages in Chinese understanding and generation, while K3 also offers excellent capabilities for English, code, and multilingual tasks, making it suitable for applications serving global users.

How does pricing work? ▾

Billing is based on Token usage, with input Tokens and output Tokens priced separately, and prices varying by model. There are no subscription or monthly fees; you pay only for what you use. Please refer to the pricing tab on this page for specific prices.

Can I use GPT, Claude, Gemini, and Kimi at the same time? ▾

Yes! Ace Data Cloud provides multiple LLMs, including GPT, Claude, Gemini, and Kimi, through a unified OpenAI-compatible interface. Simply change the model parameter to switch between models. The API format is completely consistent, so there is no need to maintain multiple codebases. The same API Key can access all models.

Other AI Models

Explore our complete AI API suite, covering large language models, images, video, and music

✨

Gemini API

Google Gemini's full range of models—million-level context and deep reasoning

→

🤖

Claude API

Anthropic's full range of Claude models—powerful reasoning and conversational capabilities

→

🎨

Flux API

Black Forest Labs' leading text-to-image model

→

🎵

Suno API

AI music generation—create custom songs and instrumental music via API

→

Start Using the Kimi K3 API Today

Use Kimi's flagship model and Max reasoning capabilities through an OpenAI-compatible format. Pay as you go, with no subscription fees.

📄 View Documentation → 💰 View Pricing

Kimi

Kimi K3 API Flagship Model for Programming & Agents