GLM Coding Plan, Tier by Tier: Lite Is the Small Plan, Not the Cheap One

The GLM Coding Plan is still one of the cheapest ways into a near-frontier coding model with a published meter. It’s also the plan people most often buy at the wrong tier—and the one most reviews still price wrong.

Quick Answer

• Worth it if you’d spend more than the plan’s price on GLM-class models anyway, and your weeks are even enough to use what you paid for.
• Not worth it if your month has one heavy week, or you need Claude or GPT in the same subscription.
• Price now: Lite $18, Pro $80, Max $168 a month. Reviews quoting $10 or less are out of date, or are referring to Chinese prices.
• Best value per credit: Max. Lite pays 50% more per credit than Max.

Prices from z.ai/subscribe, re-checked 2026-10-30.

How Much Is the GLM Coding Plan? Every Tier, Priced per Credit

Every step up buys credits more cheaply until Max, where the curve flattens. Lite isn’t the cheap plan. It’s the small plan, unlike Command Code Go or TokenHarbor Agent Pass.

The page also shows a 30% discount on yearly plans and 20% off on quarterly plans.

Did the GLM Coding Plan Price Go Up?

Yes, more than once.

That’s why so many reviews still say $10. Price the plan you’re about to buy, not the one a review priced in spring.

What Are the GLM Coding Plan Limits?

The off-peak discount is a lever you can pull every day. The model lock is a wall you can’t.

Is the GLM Coding Plan Cheaper Than the API?

Probably, by a lot—but it’s an estimate. Z.ai doesn’t publish a dollar conversion. A third-party calculation cited in our catalog puts the monthly API-equivalent value at:

Z.ai’s docs now add a token estimate of their own: 48–104 million GLM-5.3 tokens a week on Lite, at a 95–98% cache hit rate, the top of that range only if you stay off-peak.

Treat all of these as estimates, not measurements. The direction holds up: the credits sell below token list price, off-peak halves consumption, and the hard weekly cap is what lets Z.ai subsidise the plan without losing control of the bill.

What About Quality?

The top Reddit thread for this question argues the problem isn’t price, it’s output quality. We haven’t tested GLM output against Claude or GPT, so this page answers the price half only. If quality is your worry, run a week on Lite before buying a bigger tier.

Which GLM Tier Should You Buy? The Five-Minute Test

1. Take your last four weeks of coding-agent use on a comparable model.
2. Convert it to credits with the published formula (or answer three questions in the AI subscription finder).
3. Read the result:

One rule for every hard-capped plan: a cap you hit on day 12 costs more than a bigger plan you don’t. A week of waiting never shows up in a price table.

How This Was Checked

• Prices, credits, formula: z.ai/subscribe and docs.z.ai/devpack/overview, read 2026-10-30, matching our plan catalog.
• Price history: Z.ai’s announcement on X (11 February 2026), Z.ai’s legacy-plan migration notice (21 April 2026), and UsagePricing’s dated capture of z.ai (22 July 2026).
• Cost per 1,000 credits = price ÷ (weekly credits × 52 ÷ 12 ÷ 1,000), our arithmetic.
• API-equivalent value is a third-party estimate (kingy.ai, August 2026), reproduced with its range intact.
• Refunds: docs.z.ai/devpack/usage-policy, read 2026-10-30.
• Not tested hands-on. Prices move; confirm at checkout.

Buy the tier you’ll finish, not the tier that looks affordable on the first of the month.