Skip to main content

​
Overview

The Zhipu AI provider enables access to GLM (General Language Model) series from Zhipu AI (智谱AI). These models are optimized for Chinese language tasks and provide excellent performance for both Chinese and English.

​
Configuration

​
Model List Format

Add Zhipu models to your model_list configuration: 
{
  "model_list": [
    {
      "model_name": "glm-4.7",
      "model": "zhipu/glm-4.7",
      "api_key": "your-zhipu-key",
      "api_base": "https://open.bigmodel.cn/api/paas/v4",
      "request_timeout": 300
    }
  ],
  "agents": {
    "defaults": {
      "model_name": "glm-4.7"
    }
  }
}

​
Configuration Parameters

ParameterTypeRequiredDefaultDescription
model_namestringYes-Alias for this model configuration
modelstringYes-Model identifier with zhipu/ prefix
api_keystringYes-Your Zhipu AI API key
api_basestringNohttps://open.bigmodel.cn/api/paas/v4API endpoint URL
request_timeoutintegerNo120Request timeout in seconds

​
Available Models

Zhipu AI provides several GLM model families:

​
GLM-4 Series (Latest)

  • zhipu/glm-4.7 - Latest GLM-4 model, best performance
  • zhipu/glm-4 - GLM-4 base model
  • zhipu/glm-4-plus - Enhanced GLM-4 with improved capabilities
  • zhipu/glm-4-air - Lightweight GLM-4 variant
  • zhipu/glm-4-flash - Fast inference variant

​
GLM-3 Series

  • zhipu/glm-3-turbo - Fast GLM-3 model

​
Specialized Models

  • zhipu/glm-4v - Vision-capable model (multimodal)
  • zhipu/glm-4-alltools - Model with enhanced tool-calling capabilities
GLM-4 models use max_completion_tokens instead of max_tokens. PicoClaw handles this automatically.

​
Setup Instructions

​
1. Get API Key

  1. Visit Zhipu AI Platform
  2. Sign in or create an account (可使用微信登录)
  3. Navigate to API密钥管理 (API Key Management)
  4. Click 创建新的APIKey (Create New API Key)
  5. Copy your API key

​
2. Configure PicoClaw

Edit ~/.picoclaw/config.json: 
{
  "model_list": [
    {
      "model_name": "glm-4.7",
      "model": "zhipu/glm-4.7",
      "api_key": "your-actual-key-here"
    }
  ],
  "agents": {
    "defaults": {
      "model_name": "glm-4.7",
      "max_tokens": 8192,
      "temperature": 0.7
    }
  }
}

​
3. Test Connection

picoclaw agent -m "你好，测试智谱连接"

​
Advanced Configuration

​
Custom API Endpoint

Use a custom endpoint (e.g., for enterprise deployments): 
{
  "model_name": "glm-enterprise",
  "model": "zhipu/glm-4.7",
  "api_base": "https://enterprise.bigmodel.cn/api/paas/v4",
  "api_key": "your-key",
  "request_timeout": 300
}

​
Load Balancing

Configure multiple API keys for load balancing: 
{
  "model_list": [
    {
      "model_name": "glm-4.7",
      "model": "zhipu/glm-4.7",
      "api_key": "key-1"
    },
    {
      "model_name": "glm-4.7",
      "model": "zhipu/glm-4.7",
      "api_key": "key-2"
    }
  ]
}
PicoClaw automatically round-robins between endpoints with the same model_name.

​
Multimodal (Vision) Support

Use GLM-4V for image understanding: 
{
  "model_name": "glm-vision",
  "model": "zhipu/glm-4v",
  "api_key": "your-key"
}

​
Protocol Details

Zhipu AI uses OpenAI-compatible API protocol:
  • Endpoint: /chat/completions
  • Authentication: Bearer token via Authorization header
  • Request format: OpenAI-compatible JSON
  • Response format: OpenAI-compatible JSON
PicoClaw automatically:
  • Strips zhipu/ prefix when sending requests
  • Uses max_completion_tokens for GLM-4 models
  • Handles tool calling in OpenAI format

​
Troubleshooting

​
Content Filtering

Zhipu AI has content safety filters. If you receive filtering errors:
  • Rephrase your query to avoid sensitive topics
  • Use different wording
  • Check Zhipu’s content policy guidelines

​
Rate Limiting

Free tier has rate limits:
  • Configure multiple API keys for load balancing
  • Upgrade to paid plan for higher limits
  • Implement request throttling

​
Invalid API Key

Error: 401 Unauthorized
  • Verify your API key is correct
  • Check key hasn’t expired
  • Ensure sufficient credits in your account

​
Timeout Errors

Increase timeout for complex requests: 
{
  "model_name": "glm-4.7",
  "model": "zhipu/glm-4.7",
  "api_key": "your-key",
  "request_timeout": 600
}

​
Model Selection Guide

Use CaseRecommended ModelNotes
Chinese tasksglm-4.7Best for Chinese language
General tasksglm-4-plusEnhanced capabilities
Fast responsesglm-4-flashOptimized for speed
Vision tasksglm-4vSupports image input
Tool callingglm-4-alltoolsEnhanced function calling
Cost-sensitiveglm-4-airLightweight and economical

​
Cost Optimization

  1. Choose appropriate models: Use glm-4-air for simple tasks, glm-4.7 for complex tasks
  2. Set max_tokens: Limit response length to reduce costs
  3. Monitor usage: Check Zhipu Console regularly
  4. Use load balancing: Distribute requests across multiple free-tier keys
  5. Leverage free tier: 200K tokens/month free

​
Free Tier

Zhipu AI offers generous free tier:
  • 200K tokens per month for free
  • No credit card required
  • Ideal for personal projects and testing

​
Language Support

GLM models excel at:
  • Chinese: Native language, best performance
  • English: Good performance for English tasks
  • Code: Strong coding capabilities
  • Mixed: Handles Chinese-English mixed content well

​
Legacy Configuration (Deprecated)

Older configuration format (still supported): 
{
  "providers": {
    "zhipu": {
      "api_key": "your-key",
      "api_base": "https://open.bigmodel.cn/api/paas/v4"
    }
  },
  "agents": {
    "defaults": {
      "provider": "zhipu",
      "model": "glm-4.7"
    }
  }
}
Migrate to model_list format for better features like load balancing and fallbacks.

​
Example Configuration

Complete example with fallback: 
{
  "model_list": [
    {
      "model_name": "gpt4",
      "model": "openai/gpt-5.2",
      "api_key": "sk-..."
    },
    {
      "model_name": "glm",
      "model": "zhipu/glm-4.7",
      "api_key": "your-zhipu-key"
    }
  ],
  "agents": {
    "defaults": {
      "model_name": "glm",
      "max_tokens": 8192,
      "temperature": 0.7
    }
  }
}

Build docs developers (and LLMs) love

Get started for freeTalk to us