Z.ai GLM Coding Plan - Complete Reference

Last Updated: February 21, 2026 Provider: Zhipu AI (Z.ai / BigModel.cn) Documentation Sources: z.ai | docs.z.ai | open.bigmodel.cn

Related:

• Z.ai Knowledge Base - Comprehensive documentation with model specs and integrations
• Model Routing - Configuring tier-based routing
• Configuration Guide - Setting up API keys and baseUrl
• Getting Started - Quick setup instructions

Looking for more details? See the Z.ai Knowledge Base for comprehensive documentation including model specs, integrations, known issues, and historical changes.

Quick Summary

Z.ai’s GLM Coding Plan is a subscription-based AI coding service marketed as “1/7 the cost of Claude, 3x the usage”. It uses Zhipu AI’s GLM models (GLM-4.7, GLM-5) and is compatible with major coding tools like Claude Code, Cursor, Cline, and Roo Code.

Table of Contents

1. Subscription Tiers & Pricing
2. Usage Limits & Quotas
3. Model Access by Tier
4. API Endpoints & Authentication
5. Concurrency & Rate Limits
6. MCP Tool Limits
7. Integration with Coding Tools
8. Recent Changes (2026)
9. Comparison with Competitors
10. Troubleshooting & Known Issues

Subscription Tiers & Pricing

International Pricing (USD) - Post February 12, 2026

China Pricing (CNY) - Post February 12, 2026

Price Increase Notice

Effective February 12, 2026: Z.ai increased prices by 30%+ and cancelled first-purchase discounts. Existing subscribers maintain their original pricing (grandfathered in).

Usage Limits & Quotas

5-Hour Rolling Window Limits

The core quota system is based on a 5-hour rolling window, not fixed reset times. Quota dynamically recovers as requests age out of the 5-hour window.

Important Notes

• One prompt ≠ One API call: Each user prompt triggers approximately 15-20 model calls in the backend for complex coding tasks
• Rolling window: Quota used 5+ hours ago automatically releases
• No hard weekly/monthly caps: Unlike some competitors, Z.ai focuses on the 5-hour window without additional hard caps
• GLM-5 consumes more: Using GLM-5 uses significantly more quota than GLM-4.7

MCP Tool Quotas (Monthly)

Model Access by Tier

Model Specifications

Tier Model Access

GLM-5 Details

• Parameters: 754B (more than 2x GLM-4.7)
• Performance: Approaching Claude Opus 4.5 level
• Token Consumption: Significantly higher than GLM-4.7
• Launch Timeline: Pro/Max get immediate access; Lite support “coming later”

API Endpoints & Authentication

Base URLs

Coding Plan Specific Endpoint

Coding Plan subscribers must use a different endpoint than standard API users:

/api/coding/paas/v4

Authentication Methods

Bearer Token (Direct API Key)

curl -X POST "https://api.z.ai/api/paas/v4/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "glm-4.7-flash",
    "messages": [
      {"role": "user", "content": "Write a Python function to parse CSV files"}
    ],
    "max_tokens": 4096
  }'

Environment Variables for Claude Code

macOS/Linux:

export ANTHROPIC_BASE_URL="https://open.bigmodel.cn/api/anthropic"
export ANTHROPIC_API_KEY="YOUR_ZAI_API_KEY"

Windows PowerShell:

$env:ANTHROPIC_BASE_URL = "https://open.bigmodel.cn/api/anthropic"
$env:ANTHROPIC_API_KEY = "YOUR_ZAI_API_KEY"

Windows CMD:

set ANTHROPIC_BASE_URL=https://open.bigmodel.cn/api/anthropic
set ANTHROPIC_API_KEY=YOUR_ZAI_API_KEY

Python SDK

from openai import OpenAI

client = OpenAI(
    api_key="your-zai-api-key",
    base_url="https://api.z.ai/api/paas/v4/",
)

completion = client.chat.completions.create(
    model="glm-4.7-flash",
    messages=[{"role": "user", "content": "Hello, write code"}],
)

Key Management

• International API Keys: z.ai/manage-apikey/apikey-list
• Subscription Management: z.ai/manage-apikey/subscription
• China Platform: open.bigmodel.cn

Concurrency & Rate Limits

Official Documentation Status

Important: Zhipu AI does NOT publicly document specific TPM (Tokens Per Minute) or RPM (Requests Per Minute) rate limits for GLM Coding Plan subscriptions. Their documentation focuses on “prompts per 5 hours” rather than per-minute limits.

Known Concurrency Limits

Peak Hour Restrictions

Known Issues (as of January 2026):

• Peak Hours: Weekdays 15:00-18:00 (3-6 PM)
• Symptoms: Concurrent rate limiting errors, slower response times
• Root Cause: “Computing resource phase tension” due to high demand

Official Capacity Measures

Since January 23, 2026:

• Daily sales limited to 20% of capacity
• Daily quota refreshes at 10:00
• Auto-renewal subscribers unaffected
• Active crackdown on malicious traffic and violations

Tier Concurrency Differences

Recommended Usage Strategy

1. Avoid peak hours when possible (15:00-18:00 weekdays)
2. Use Max tier for production/high-concurrency needs
3. Enable auto-renewal for more stable resource allocation
4. Monitor usage with quota tracking tools like opencode-glm-quota

MCP Tool Limits

MCP Tools Available

• Image Analysis (vision)
• Web Search (web_search)
• Web Reader (web_reader)
• Open Source Repository MCP

Monthly MCP Quotas

MCP Usage Tracking

The 24-hour rolling window tracks:

• Model usage (prompts)
• MCP tool usage (web_search, web_reader, vision)

Use the opencode-glm-quota tool for real-time monitoring.

Integration with Coding Tools

Supported Tools

Z.ai GLM Coding Plan is compatible with 10+ mainstream coding tools:

Claude Code Integration

Step 1: Set Environment Variables

# For Anthropic-compatible endpoint
export ANTHROPIC_BASE_URL="https://open.bigmodel.cn/api/anthropic"
export ANTHROPIC_API_KEY="your-zai-coding-plan-api-key"

Step 2: Configure Model

In Claude Code settings or .claude/settings.json:

{
  "provider": "anthropic",
  "model": "glm-4.7"
}

Cursor Integration

Settings → API Providers → Custom:

{
  "baseURL": "https://api.z.ai/api/paas/v4/",
  "apiKey": "your-api-key",
  "model": "glm-4.7"
}

Cline (VS Code Extension)

Settings JSON:

{
  "cline.apiUrl": "https://api.z.ai/api/paas/v4/",
  "cline.apiKey": "your-api-key",
  "cline.model": "glm-4.7"
}

Recent Changes (2026)

February 12, 2026 - Price Increase

GLM-5 Launch

Capacity Constraints

Since January 2026:

• 2.7 million paid users (as of December 2025)
• Daily sales limited to 20% of capacity
• Peak hour rate limiting (15:00-18:00 weekdays)
• Aggressive expansion of computing infrastructure underway

Comparison with Competitors

Value Proposition

Market Positioning

Z.ai claims “21x better value than Claude”:

• 1/7 of Claude’s price (at equivalent usage)
• 3x the usage quota
• Comparable model performance (GLM-4.7 ≈ Claude Sonnet 4)

Chinese Market Alternatives

*First-month promotional pricing

Troubleshooting & Known Issues

Common Error Messages

Known Issues

1. Undocumented Concurrency Limit (GitHub Issue)

• Issue: GLM 4.7 appears to only allow 1 concurrent in-flight request, even on paid tiers
• Status: Unconfirmed officially, reported by users
• Workaround: Implement queue-based request handling

2. Peak Hour Slowdowns

• Issue: 15:00-18:00 weekdays experience significant slowdowns
• Status: Officially acknowledged
• Workaround: Schedule heavy work for off-peak hours or upgrade to Max tier

3. GLM-5 High Token Consumption

• Issue: GLM-5 uses significantly more quota per prompt than GLM-4.7
• Status: By design (larger model)
• Workaround: Use GLM-4.7 for routine tasks, GLM-5 for complex reasoning

Getting Help

• China Documentation: docs.bigmodel.cn/cn/coding-plan/overview
• International Docs: docs.z.ai/devpack/overview
• Quick Start: docs.z.ai/devpack/quick-start
• GitHub Community: Search for zai-coding or glm-coding-plan

SDKs & Libraries

Official SDKs

Monitoring Tools

• opencode-glm-quota - Real-time quota monitoring for GLM Coding Plan
  • 5-hour token cycle tracking
  • Monthly MCP usage statistics
  • 24-hour rolling window for model and MCP tools

Quick Start Checklist

• Create account at z.ai (international) or open.bigmodel.cn (China)
• Generate API key from API management console
• Subscribe to GLM Coding Plan tier (Lite/Pro/Max)
• Configure environment variables for your coding tool
• Test API connection with a simple curl request
• Monitor usage with quota tracking tool
• Adjust usage patterns based on 5-hour rolling window

Cross-References

Internal Documentation

• Model Mapping - How Z.ai models map to internal model IDs
• Model Routing - Request routing logic
• Model Concurrency Findings - Concurrency testing results

External Documentation

• Z.ai Official Docs
• BigModel China Docs
• Claude Code Setup

This document is maintained as a living reference. Last comprehensive search: February 21, 2026.