The GLM Coding Plan gives you Z.ai's GLM-5.2 model inside tools like Claude Code. There are three tiers, Lite, Pro and Max, and here is how to choose between them, plus how to take 5% off whichever one you pick.

Why we don't list a price here

Two reasons, both honest:

  1. Region matters a lot. The overseas GLM Coding Plan costs noticeably more than the mainland-China one (often around 3x). Any single number we printed would be wrong for half our readers.
  2. It changes. Z.ai adjusts pricing and runs promotions regularly, and the effective cost also depends on whether you pay monthly or commit to a longer billing cycle.

So instead of a figure that goes stale, check the live price for your region on the Z.ai subscribe page. Whatever it shows, our referral code takes 5% off your first order.

The three tiers at a glance

The tiers differ by usage quota, not by which models you get. Every tier includes the same models.

Tier Prompts per 5h Prompts per week Monthly MCP calls Best for
Lite up to ~80 up to ~400 100 Solo devs, daily sessions
Pro up to ~400 up to ~2,000 1,000 Power users, agentic work
Max up to ~1,600 up to ~8,000 4,000 All-day, multi-repo heavy use

Lite: the entry tier

Lite is the sweet spot for most individual developers. You get up to ~80 prompts per rolling 5-hour window (about 400 a week) plus 100 monthly web search / reader MCP calls. If you do one or two focused coding sessions a day, Lite is usually enough, with full access to GLM-5.2 and its 1M-token context.

Pro: for power users

Pro steps up to roughly 5x Lite's usage (up to ~400 prompts per 5 hours, ~2,000 a week) and 1,000 monthly MCP calls. This is the tier to pick if you are running agentic, multi-file tasks throughout the day.

Max: for heavy, all-day use

Max is built for developers who live in their coding agent: the highest quota (up to ~1,600 prompts per 5 hours, ~8,000 a week) and 4,000 monthly MCP calls. If you are hitting limits on Pro, Max removes the ceiling for serious daily workloads.

What every tier includes

All three tiers get the same models, not a cut-down version on the cheaper plans:

  • GLM-5.2, the flagship coding-first model with a 1M-token context window
  • GLM-5-Turbo, a faster premium model
  • GLM-4.7, handy for lighter, routine tasks to conserve quota
  • Four bundled MCP tools: Vision Understanding, Web Search, Web Reader and Zread

The tiers differ in quota, not model access.

How the quota actually works

Usage is measured in prompts on a rolling 5-hour window. As time passes, your oldest usage falls off and quota frees up automatically; there is no fixed daily reset. One "prompt" is one query you send, but Z.ai notes each prompt invokes the model roughly 15 to 20 times behind the scenes in an agentic tool, since it searches, reads, edits and validates code.

One more nuance: the advanced models (GLM-5.2 and GLM-5-Turbo) consume quota faster than the lighter ones, normally counting as more than one unit per call during peak hours. A limited-time promotion is billing them at the off-peak rate through around September 2026. If you want to stretch quota, use GLM-4.7 for routine work and save GLM-5.2 for the hard tasks.

Tips for paying less

  • Commit to a longer billing cycle. Paying for a quarter or a year up front lowers the effective monthly rate compared with month-to-month.
  • Apply our referral code. Code RIMBTGLNJI (or the referral link) takes 5% off your first order on any tier. It is first-order only and not stackable.

Which tier should you pick?

  • Just trying it or light use goes to Lite
  • Daily agentic coding goes to Pro
  • Heavy, professional all-day use goes to Max

Start on Lite or Pro; you can upgrade later if you hit limits. See our GLM Coding Plan vs Claude Code comparison to decide if it is right for your workflow, then grab 5% off.