Kimi K3

kimi-k3

Open sourceKimi series

Overview

1M ContextDocument analysisReasoningAgentic tasks

Kimi K3 is Moonshot AI's flagship multimodal model, built around a 1M-token context window and tiered deep reasoning. It keeps hold of the goal through long-document reading, cross-chapter references and multi-turn tool calls; combined with function calling, MCP tools and structured output, it can drive enterprise knowledge-base Q&A and complex automation pipelines.

1M-token context

Loads an entire technical manual, a long research report or a whole code repository in a single pass, with no drop-off in cross-chapter retrieval.

Tiered deep reasoning

Two effort levels, high and xhigh, schedule compute on demand to balance cost against quality.

Structured output

Native JSON Schema and function call support plugs straight into business systems and databases.

Open weights, self-hostable

The K3 series has open weights, so the model can be customized and integrated deeply.

Features

Tool calling✓

Structured output✓

Context caching✓

Batch✓

Web search✓

Streaming✓

Reasoning✓

MCP✓

Prefix completion✕

Fine-tuning✕

Pricing

Input¥20/M tokens

Input (cache hit)¥2/M tokens

Output¥100/M tokens

Rate Limits & Context

Context Window1M

Max Output1M

RPM (requests/min)500

TPM (tokens/min)3M

Built-in Tools

web_search

Responses API

code_interpreter

Responses API

web_extractor

Responses API

API Reference

ENDPOINT

POST https://api.tokenfab.cn/v1/chat/completions

Request Example

curl https://api.tokenfab.cn/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKENFAB_API_KEY" \
  -d '{
    "model": "kimi-k3",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "请总结这份 30 万字的技术文档，并提炼 5 个核心结论。"}
    ]
  }'

from openai import OpenAI

client = OpenAI(
    api_key="your-api-key",
    base_url="https://api.tokenfab.cn/v1"
)

response = client.chat.completions.create(
    model="kimi-k3",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "请总结这份 30 万字的技术文档，并提炼 5 个核心结论。"}
    ]
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.TOKENFAB_API_KEY,
  baseURL: "https://api.tokenfab.cn/v1"
});

const response = await client.chat.completions.create({
  model: "kimi-k3",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "请总结这份 30 万字的技术文档，并提炼 5 个核心结论。" }
  ]
});

console.log(response.choices[0].message.content);

Response Example

JSON

{
  "id": "chatcmpl-kimi-k3-9d2k7f1",
  "object": "chat.completion",
  "created": 1784950000,
  "model": "kimi-k3",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "本文核心结论：(1) 长上下文窗口是新一代 LLM 的关键差异；(2) 推理与工具调用协同提升复杂任务表现；(3) 缓存显著降低长对话成本；(4) 结构化输出对接业务系统效率倍增；(5) 多模态能力与开源生态推动模型成为通用智能基础设施。"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 287412,
    "completion_tokens": 96,
    "total_tokens": 287508
  }
}

Request Parameters

Parameter	Type	Required	Description
model	string	✓	Model ID, e.g. kimi-k3
messages	array	✓	List of chat messages, each with role and content
temperature	float	—	Sampling temperature, 0–2, default 0.6
top_p	float	—	Nucleus sampling, 0–1, default 0.9
max_tokens	int	—	Maximum output tokens
stream	bool	—	Whether to stream the response, default false
tools	array	—	Tool definitions for function calling
tool_choice	string	—	Tool choice policy: auto / none / a named function