概述
Kimi K3 是 Moonshot AI 月之暗面推出的旗舰多模态模型,主打 1M 超长上下文与多档深度推理。在长文档阅读、跨章节引用、多轮工具调用等场景中持续保持目标一致性,配合 Function Calling、MCP 工具与结构化输出,可直接驱动企业级知识库问答与复杂自动化流水线。
1M 超长上下文
支持整本技术手册、长篇研报、整仓库代码一次性载入,跨章节检索无衰减。
多档深度推理
提供 high / xhigh 双档推理强度,按需调度算力,兼顾成本与质量。
结构化输出
原生支持 JSON Schema 与 function call,无缝对接业务系统与数据库。
开源可部署
K3 系列权重开源,模型能力可定制与深度集成。
功能
工具调用✓
结构化输出✓
上下文缓存✓
批量任务✓
联网搜索✓
流式输出✓
深度思考✓
MCP✓
前缀补全✕
微调✕
定价
输入$3/M tokens
输入(缓存命中)$0.3/M tokens
输出$15/M tokens
速率限制与上下文
上下文窗口1M
最大输出1M
RPM(请求 / 分钟)500
TPM(token / 分钟)3M
内置工具
code_interpreter
Responses API
web_extractor
Responses API
web_search
Responses API
API 参考
POST
https://api.tokenfab.com/v1/chat/completions请求示例
curl https://api.tokenfab.com/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": "Summarize this 300,000-word technical document and distill 5 key conclusions."}
]
}'import os
from openai import OpenAI
client = OpenAI(
api_key=os.getenv("TOKENFAB_API_KEY"),
base_url="https://api.tokenfab.com/v1"
)
response = client.chat.completions.create(
model="kimi-k3",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Summarize this 300,000-word technical document and distill 5 key conclusions."}
]
)
print(response.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.TOKENFAB_API_KEY,
baseURL: "https://api.tokenfab.com/v1"
});
const response = await client.chat.completions.create({
model: "kimi-k3",
messages: [
{ role: "system", content: "You are a helpful assistant." },
{ role: "user", content: "Summarize this 300,000-word technical document and distill 5 key conclusions." }
]
});
console.log(response.choices[0].message.content);响应示例
{
"id": "chatcmpl-kimi-k3-9d2k7f1",
"object": "chat.completion",
"created": 1784950000,
"model": "kimi-k3",
"choices": [{
"index": 0,
"message": {
"role": "assistant",
"content": "Key conclusions: (1) long context windows are the key differentiator of new-generation LLMs; (2) reasoning combined with tool calling improves performance on complex tasks; (3) caching significantly lowers the cost of long conversations; (4) structured output makes integration with business systems far more efficient; (5) multimodal capabilities and the open-source ecosystem are turning models into general-purpose AI infrastructure."
},
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": 287412,
"completion_tokens": 96,
"total_tokens": 287508
}
}请求参数
| 参数 | 类型 | 必填 | 说明 |
|---|---|---|---|
model | string | ✓ | 模型 ID,如 kimi-k3 |
messages | array | ✓ | 对话消息列表,含 role 和 content |
temperature | float | — | 采样温度,0–2,默认 0.6 |
top_p | float | — | 核采样,0–1,默认 0.9 |
max_tokens | int | — | 最大输出 Token 数 |
stream | bool | — | 是否流式输出,默认 false |
tools | array | — | 工具定义列表,用于 Function Calling |
tool_choice | string | — | 工具调用策略,auto / none / 指定函数 |