Apr 25, 2026 · 4 min read

Last updated on Apr 19, 2026

Z.ai API Complete Guide — GLM Models, Pricing, and Setup (2026)

Z.ai (formerly Zhipu AI) offers API access to the GLM model family through their Coding Plan. The standout feature: an Anthropic-compatible endpoint that lets you use GLM-5.1 through Claude Code.

Available models

Model	Quality	Speed	Best for
GLM-5.1	Best	Medium	Complex coding, architecture
GLM-5-Turbo	Very good	Fast	Balanced quality/speed
GLM-4.7	Good	Fast	Routine coding, budget
GLM-4.5-Air	Basic	Fastest	Quick questions, autocomplete

All models are included in the Coding Plan — no per-model pricing.

Pricing: the Coding Plan

Plan	Price	Quota	Models
Coding Plan Lite	$18/month	5-hour blocks + weekly quota	All 4 models

How quota works

All models consume from the same quota pool:

Model	Off-peak rate	Peak rate (14:00-18:00 UTC+8)
GLM-5.1	1x	2-3x
GLM-5-Turbo	1x	2-3x
GLM-4.7	1x	1x
GLM-4.5-Air	1x	1x

Key insight: During off-peak hours, GLM-5.1 costs the same as GLM-4.7. For European and US developers, peak hours (14:00-18:00 UTC+8 = 08:00-12:00 CEST = 02:00-06:00 ET) are early morning — you’re likely sleeping anyway.

See our GLM-5.1 API pricing guide for detailed quota analysis.

Setup: Anthropic-compatible endpoint

The killer feature of Z.ai is the Anthropic-compatible API. This means any tool that works with Claude also works with GLM:

export ANTHROPIC_AUTH_TOKEN="your-zai-api-key"
export ANTHROPIC_BASE_URL="https://api.z.ai/api/anthropic"

With Claude Code

export ANTHROPIC_AUTH_TOKEN="your-zai-api-key"
export ANTHROPIC_BASE_URL="https://api.z.ai/api/anthropic"
export ANTHROPIC_DEFAULT_OPUS_MODEL="glm-5.1"
export ANTHROPIC_DEFAULT_SONNET_MODEL="glm-4.7"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="glm-4.5-air"

claude  # Now uses GLM-5.1 instead of Claude

You get the full Claude Code agentic experience at $18/month. See our complete Claude Code + GLM setup guide.

With Python (OpenAI-compatible)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.z.ai/api/openai/v1",
    api_key="your-zai-api-key",
)

response = client.chat.completions.create(
    model="glm-5.1",
    messages=[{"role": "user", "content": "Review this code for bugs"}],
)

With Aider

export OPENAI_API_BASE="https://api.z.ai/api/openai/v1"
export OPENAI_API_KEY="your-zai-api-key"
aider --model glm-5.1

Getting your API key

Sign up at z.ai
Subscribe to the Coding Plan Lite ($18/month)
Go to API Keys in your dashboard
Generate a new key
Use it with the endpoints above

Quota management tips

From our AI Startup Race testing:

A 30-minute GLM-5.1 session uses ~44% of the 5-hour quota block
A 30-minute GLM-4.7 session uses ~35% of the 5-hour quota block
Weekly quota of 22% per premium session means ~4-5 GLM-5.1 sessions per week

Sustainable daily schedule: 1 premium (GLM-5.1) + 1 cheap (GLM-4.7) session per day.

Maximize value:

Use GLM-5.1 for complex tasks (architecture, debugging, code review)
Use GLM-4.7 for routine tasks (refactoring, simple generation)
Schedule GLM-5.1 sessions during off-peak hours (1x rate instead of 2-3x)
Monitor quota in the Z.ai dashboard

Z.ai vs other API providers

Provider	Best model	Monthly cost	Anthropic-compatible
Z.ai	GLM-5.1	$18	✅
Anthropic	Claude Sonnet	$20	✅ (native)
OpenRouter	Multiple	Pay per token	❌ (OpenAI-compatible)
DeepSeek	DeepSeek V3	~$25	❌
Moonshot	Kimi K2.5	~$19	❌

Z.ai is the only provider besides Anthropic that offers an Anthropic-compatible endpoint. This makes it the only drop-in replacement for Claude Code.

Rate limits and reliability

Rate limits: Included in the Coding Plan quota, no separate rate limit
Uptime: Generally reliable, occasional slowdowns during Chinese business hours
Regions: Servers primarily in China, ~100-200ms latency from Europe/US
Support: Chinese-language support primarily, English documentation available

For production use outside of coding tools, consider the latency. For Claude Code sessions where you’re waiting for responses anyway, the latency is not noticeable.

FAQ

What is Z.ai?

Z.ai is the international-facing AI platform from Zhipu AI, a leading Chinese AI company. It provides API access to the GLM model family, including GLM-5.1, GLM-5-Turbo, and GLM-4.7, through both OpenAI-compatible and Anthropic-compatible endpoints.

Is Z.ai the same as Zhipu AI?

Z.ai is the global brand and API platform operated by Zhipu AI. Zhipu AI is the parent company that develops the GLM models, while z.ai is the domain and product name used for their international developer-facing services and Coding Plan subscriptions.

How much does Z.ai cost?

The Coding Plan Lite is $18/month and includes access to all GLM models (GLM-5.1, GLM-5-Turbo, GLM-4.7, GLM-4.5-Air) with a quota-based system. There is no pay-per-token option through Z.ai directly, though you can access GLM models via OpenRouter on a per-token basis.

Can I use Z.ai with Claude Code?

Yes — this is Z.ai’s killer feature. The Anthropic-compatible endpoint at https://api.z.ai/api/anthropic lets you use GLM-5.1 as a drop-in replacement for Claude in Claude Code. Just set the ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN environment variables to point to Z.ai.