GPT88 Agent Docs
ChatMoonshotFeatured1 vendors

kimi-k3

Moonshot flagship model for long-horizon coding, knowledge work, and complex agent tasks.

Get API Key on gpt88.ccQuickstart

Overview

  • Kimi K3 is a flagship model available through Moonshot’s Kimi API Platform. Its model ID is kimi-k3.
  • Moonshot describes a 2.8T-parameter architecture with native visual understanding and a 1M-token context window, aimed at long-horizon coding, knowledge work, and reasoning.

Best Use Cases

  • Understanding large codebases and carrying out multi-step engineering work
  • Processing long documents, research material, and structured knowledge tasks
  • Evaluating visual inputs, tool calls, or agent workflows

Integration Notes

  • Use the GPT88 Base URL https://api.gpt88.cc, model ID kimi-k3, and endpoint /v1/chat/completions.
  • Kimi K3 works with OpenAI-compatible SDKs, cURL, and agent tools. Call GET /v1/models first to confirm access for the current API key.
  • Moonshot’s direct API documentation uses https://api.moonshot.ai/v1. GPT88 integrations should use the unified Base URL published by GPT88.

Usage Notes

  • Model availability, routes, pricing, rate limits, and permissions are determined by the current GPT88 console configuration.
  • A 1M-token context window does not mean every request should send the full context. Split work by task, reuse cached context, and monitor latency and cost.
  • Validate the model on representative codebases, long documents, and tool-calling workloads before a production rollout.

Capabilities

1M-token contextLong-horizon codingKnowledge workNative visionTool calling

Recommended Scenarios

  • Large codebase analysis
  • Long documents and research
  • Complex reasoning
  • Multi-step agents

Endpoint Path

POSThttps://api.gpt88.cc/v1/chat/completions

The endpoint path is determined by the model category. Chat, image, video, and audio models use different endpoints.

Send requests with Authorization: Bearer <API_KEY>. See Chat Completions API for the full parameter reference.

API Docs

Endpoint
POST /v1/chat/completions
Model ID
kimi-k3
Content-Type
application/json
Protocol
OpenAI-compatible chat completions endpoint with streaming and non-streaming output.
Base URL
Use https://api.gpt88.cc for OpenAI-compatible and Claude-style tools.

Headers / Auth

  • AuthorizationstringRequired
    Send Bearer <GPT88_API_KEY>. Create the API key in the gpt88.cc console.
  • Content-TypestringRequired
    Request body format. Audio transcription uploads use multipart/form-data.
    Default:application/json
  • Acceptstring
    Non-streaming requests return JSON; streaming chat requests return SSE chunks.
    Default:application/json or text/event-stream

Chat Protocol Differences

For chat models, the protocol used by the client matters as much as the model name. Claude Code / Anthropic SDK and OpenAI SDK build different request paths.

OpenAI CompatibleRecommended
Base URL
https://api.gpt88.cc
Endpoint
POST /v1/chat/completions
Body
{ "model": "kimi-k3", "messages": [...] }
Anthropic / Claude
Base URL
https://api.gpt88.cc
Endpoint
POST /v1/messages
Body
{ "model": "kimi-k3", "messages": [...] }

This model is usually best accessed through the OpenAI-compatible protocol. Only use the Anthropic-style path when the target tool explicitly requires it.

Request Parameters

The fields below are generated by model category. Pricing, context limits, rate limits, and available routes should be confirmed in the gpt88.cc console.

  • modelstringRequired
    Current model ID: kimi-k3. Preserve the exact casing when copying.
  • messagesarray<Message>Required
    Conversation history. Each message includes role and content.
  • streamboolean
    Enable SSE streaming. Useful for chat interfaces and long responses.
    Default:false
  • temperaturenumber
    Sampling temperature. Higher values produce more variation; production workloads usually start with a fixed value.
    Default:1
  • max_tokensinteger
    Maximum output tokens for this request. Real context limits should be confirmed in the console.
  • response_formatobject
    Structured output config, for example { "type": "json_object" }.
  • toolsarray<Tool>
    Function-calling tool definitions. Availability depends on the model and account configuration.

Response Fields

Response bodies generally follow OpenAI-compatible shapes. Media models may return async task IDs or resource URLs.

  • idstringRequired
    Unique completion ID for tracing and debugging.
  • objectstringRequired
    Usually chat.completion or a streaming chunk type.
  • modelstringRequired
    Actual model ID that handled inference.
  • choicesarray<Choice>Required
    Generated choices including message, delta, and finish_reason.
  • usageobject
    Token usage statistics. Streaming calls may return this only at the end or in non-streaming mode.

Status Codes

  • 200OKRequired
    Request succeeded. See the response field table above for the response body shape.
  • 400Bad Request
    Invalid request fields, such as missing required fields, unsupported image size format, or unsupported file type.
  • 401Unauthorized
    API key is missing, invalid, or malformed.
  • 404Not Found
    Endpoint or model not found. Confirm the path is /v1/chat/completions and the model ID is kimi-k3.
  • 429Rate Limited
    Rate limit, concurrency cap, or insufficient balance. Check the console for the current allowance.
  • 5xxUpstream Error
    Upstream or request failure. Retry with the same Base URL and keep the request ID for debugging.

Errors and Troubleshooting

  • 401: verify Authorization is Bearer <API_KEY> and the key is still valid.
  • 404: confirm the endpoint is /v1/chat/completions and the model ID kimi-k3 is available in the console.
  • 429: reduce concurrency, shorten requests, or check balance and quota in the console.
  • 5xx: retry with the same Base URL and keep the request ID for debugging.

Request Examples

curl https://api.gpt88.cc/v1/chat/completions \
  -H "Authorization: Bearer $GPT88_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [
      {"role": "user", "content": "Introduce gpt88.cc in one sentence."}
    ]
  }'

Expected response (trimmed example):

200 OKjson
{
  "id": "chatcmpl-kimi-k3",
  "object": "chat.completion",
  "created": 1730000000,
  "model": "kimi-k3",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": { "role": "assistant", "content": "..." }
    }
  ],
  "usage": { "prompt_tokens": 24, "completion_tokens": 64, "total_tokens": 88 }
}