Kimi API · Ace Data Cloud

Kimi K3 API
Flagship model for programming and agents

Name: Kimi
Brand: Ace Data Cloud

Call Kimi K3 in OpenAI compatible format. 1M Token context, native visual understanding, and long-range programming capabilities, with reasoning_effort to control inference effort.

📄 View Documentation 💰 View Pricing

🚀 Kimi K3 🧠 Max Inference 🔌 OpenAI Compatible 🌐 Bilingual in Chinese and English

🤖

Flagship Model

🧠

2.8T

Total MoE Parameters

💰

Low Price

Below Official Pricing

⚡

Context Window

Why Use Kimi Through Ace Data Cloud?

Kimi K3 is the flagship model of the Dark Side of the Moon, aimed at long-range programming, complex reasoning, and end-to-end knowledge work. It has a 1M Token context window, capable of continuously understanding large codebases and coordinating tools to complete multi-step tasks.

Ace Data Cloud offers Kimi K3 and K2 Series APIs, using OpenAI compatible format. Existing OpenAI SDKs can be integrated by simply changing the base URL and model name.

Core Capabilities of Kimi K3 API

Provides a unified interface for complex programming, Agents, and knowledge work

🔌

OpenAI Compatible Format

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

🧠

Max Inference Mode

K3 always enables inference, the only supported value for reasoning_effort is max, suitable for complex programming, mathematics, and Agent tasks.

🌐

Bilingual Optimization in Chinese and English

Excellent native understanding of Chinese, while also performing well in English. Suitable for Chinese scenarios, cross-language translation, and multilingual application development.

💻

Top Programming Capability

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

⚡

Streaming

Supports SSE streaming, real-time token-by-token output. Set stream: true for a streaming response experience.

📄

Structured Output

Specify JSON output format via response_format, along with 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": "Please answer concisely, output only tested Python code, do not explain; ensure that duplicate elements do not cause infinite recursion."},
        {"role": "user", "content": "Implement a quicksort algorithm in Python"}
    ],
  reasoning_effort="max",
  stream=False
)
print(response.choices[0].message.content)

Actual Response (Simplified)

{
  "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
  }
}

OpenAI SDK One-Line Code Switch

Simply change base_url and model to use Kimi K3 in your existing OpenAI project, and adjust the reasoning input with reasoning_effort.

Get API Key

Modify base_url

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

Select Kimi Model

Set model to kimi-k3; if you need to explicitly set reasoning parameters, use reasoning_effort="max"

What can be built using Kimi API?

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

💬

Chinese Conversation Assistant

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

💻

Code Generation and Review

K3 supports code generation, bug fixing, code review, architectural suggestions, and multi-step terminal tool collaboration

🌐

Cross-Language Translation

Native Chinese-English bilingual capability, suitable for high-quality translation, localization, and multilingual content generation

🔬

Deep Reasoning and Analysis

Mathematical problem solving, logical reasoning, data analysis—Thinking mode provides a step-by-step reasoning process

Quick Start in 3 Steps

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

Register and Get API Key

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

Call Using OpenAI SDK

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

Integration and Expansion

Embed Kimi K3 into your application. OpenAI compatible format makes switching between multiple models a breeze.

View Documentation →

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

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

Comparison Dimension	Ace Data Cloud	Moonshot Direct Connection
OpenAI Compatible Format	✓	Partially Compatible
Global Availability	✓ Ready to use	✗ Limited in some regions
Streaming	✓	✓
Unified Interface for Multiple Models	✓ GPT / Claude / Gemini / Kimi	✗ Only Kimi
Pay-as-you-go	✓ Flexible top-up	✓
No overseas phone number registration required	✓	✗ Requires a Chinese phone number
Thinking Reasoning Model	✓	✓

Select the Suitable Kimi Model

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

K2.6

General Efficient

General model, supports reasoning, tool invocation, and multimodal input, suitable for daily conversations and Agent tasks.

✓ 256K Context
✓ Toggle Thinking
✓ Suitable for general conversation and analysis
✓ Supports tool invocation

Recommended

Kimi K3

Flagship Model

The flagship model for long-range programming and complex Agent workflows, reasoning is always enabled, currently supports Max reasoning mode.

✓ 1M Token Context
✓ reasoning_effort: max
✓ Tool invocation and streaming output
✓ Preferred for complex programming tasks

K2.5

Compatible

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

✓ 256K Context
✓ Toggle 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

Charged by Token usage. No subscription fees, no hidden costs.

Bulk packages enjoy more discounts

Pay-as-you-go

Token Billing

Low Price Charged by Token

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

✓ K3, K2.6, and K2.5 transparently priced by model
✓ Inference Tokens charged based on actual usage
✓ No subscription fees or minimum consumption
✓ Separate pricing for input and output
✓ Streaming — Free

View Pricing Details View API Documentation

Enterprise Edition

Custom

Exclusive plans for high-usage teams

✓ Tiered discounts based on usage
✓ Priority support and account manager
✓ Custom rate limits
✓ SLA guarantees
✓ 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 the flagship MoE model of the Dark Side of the Moon, with a total of 2.8 trillion parameters and a 1M Token context window, suitable for understanding large codebases, long-range programming, complex reasoning, tool invocation, and end-to-end knowledge work.

Does it support OpenAI SDK? ▾

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

How should reasoning_effort be set? ▾

K3 always enables reasoning, and the only currently supported value is max; if omitted, max is used as well. Some upstream compatibility may accept other strings, but it does not guarantee a change in reasoning behavior, so it is not recommended to use.

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

How does its Chinese capability compare to GPT/Claude? ▾

The Kimi series has a native advantage in understanding and generating Chinese, while K3 also possesses excellent capabilities in English, code, and multilingual tasks, suitable for applications aimed at global users.

What is the pricing model? ▾

Billing is based on Token usage, with separate pricing for input Tokens and output Tokens, and different models have different prices. There are no subscription fees or monthly fees; you pay for what you use, and the specific prices are subject to the pricing labels on this page.

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

Yes! Ace Data Cloud provides multiple LLMs such as GPT, Claude, Gemini, and Kimi through a unified OpenAI-compatible interface. You can switch between different models simply by changing the model parameter; the API format is completely consistent, and there is no need to maintain multiple sets of code. The same API Key can access all models.