GLM 5.2 Pricing: API Cost, Plans & Free Tiers

I almost paid for the wrong GLM 5.2 plan. The pricing pages send you in circles—token rates in one place, a "Coding Plan" in another, promo prices that quietly reset after launch week—and it's easy to overpay or pick the wrong access route. So I mapped every way to pay for GLM 5.2, ran the numbers, and here's the short version: GLM 5.2 costs about $1.40 per million input tokens and $4.40 per million output on Z.AI's official API—roughly a fifth to a sixth of what Claude Opus 4.8 or GPT-5.5 charge—and there are cheaper routes plus free ways to run it. Below is the full breakdown and how to pick the cheapest option for how you actually work.

GLM 5.2 Pricing at a Glance

Note: Self-hosting is "free" of per-token fees, but you pay for the hardware (the full weights need serious GPU memory). Prices reflect launch-period rates (mid-June 2026) and change often—verify current numbers on each vendor's page.

If you just want to use GLM 5.2 without thinking about any of this, the top row is your answer—glm5.app is free and needs zero setup. If you're a developer who needs metered API access, OpenRouter is the cheapest. Now let's break down each route.

GLM 5.2 API Pricing (Pay-as-You-Go)

On Z.AI's official API, GLM 5.2 is billed per token at roughly the following rates—as independently tracked by Artificial Analysis:

• Input: ~$1.40 per 1M tokens
• Output: ~$4.40 per 1M tokens
• Cached input: ~$0.26 per 1M tokens

That cached-input rate is the one people miss. If your app reuses the same long context—a system prompt, a codebase, a document—prompt caching can cut the effective input cost dramatically on repeated calls.

To make it concrete, here's a typical agentic coding turn at Z.AI's listed rates: feed in ~1M tokens of context and get back 200K tokens of output, and you're looking at roughly $1.40 + $0.88 ≈ $2.30. Lean on caching for the repeated context and that same turn drops toward **$1.15**. Your real numbers depend on how chatty your prompts are, but the order of magnitude holds: this is a model you can run hard without a frightening bill.

Cheaper Access: OpenRouter & Other Providers

You don't have to buy direct. Because GLM 5.2 ships as open weights, several providers host it, and the metered prices differ. On OpenRouter's GLM 5.2 model page, the listed rate is ~$0.95 per 1M input and ~$3.00 per 1M output—cheaper than Z.AI's own API on both sides.

The catch: third-party providers vary on throughput, context limits, and uptime, and the exact numbers shift. Treat any single quote as a snapshot and confirm it on the provider's page before you route production traffic. But for cost-sensitive, high-volume work, a router like OpenRouter is usually the cheapest pay-as-you-go path.

GLM Coding Plan: Subscriptions

If you live in an IDE or CLI all day, per-token billing is the wrong model—you want a flat monthly fee. Z.AI's GLM Coding Plan does that, with tiers running from a light entry plan up through Pro, Max, and Team seats, each raising how many prompts you get per cycle.

Two honest caveats before you subscribe:

1. Launch promos make the entry prices look incredible—and they reset. The lowest numbers floating around reflect introductory pricing; standard monthly rates settled higher after launch. Check the live price, not a launch-week screenshot.
2. GLM 5.2 burns quota faster than older models—reportedly around 2× off-peak and up to 3× at peak—so a tier's "prompts per week" goes further on paper than in heavy practice.

For exact current tier prices and allowances, check Z.AI's plan page directly; they move too often to quote reliably here.

Free Ways to Use GLM 5.2

"Is GLM 5.2 free?" has three real answers:

1. Self-host the open weights. They're published under an MIT license, so you can download and run GLM 5.2 yourself with no per-token fee—you just need the hardware (the full model wants multiple high-memory GPUs).
2. New-user and daily credits. Z.AI and partner platforms have offered new-account token grants and daily free credits. These come and go, so treat any specific token figure as a claim to verify, not a guarantee.
3. Just open a browser. The fastest free path with zero setup is glm5.app—chat with GLM 5.2 right in your browser, no install, no API key, no card. It's the easiest way to feel the model on a real task before you decide which paid route is worth it.

GLM 5.2 vs Claude vs GPT-5.5 Pricing

Here's why GLM 5.2's pricing gets attention—the same work costs a fraction of the closed frontier:

On output tokens—where coding and agent workloads spend most—GLM 5.2 runs roughly five to six times cheaper. VentureBeat pegged the blended, all-in gap at about one-sixth the cost of GPT-5.5. And it earns that price: GLM 5.2 isn't a budget compromise but a genuine frontier-class coder. If you want the receipts on performance before you trust the price, see how GLM 5.2 actually performs on benchmarks, or the full overview of what GLM 5.2 is.

Which Pricing Option Should You Pick?

Here's the honest shortcut: every paid route below charges a setup tax—API keys, billing dashboards, subscription tiers, or GPU hardware. If you just want to use GLM 5.2, you can skip all of it.

• Most people: just use it free in the browser. glm5.app runs GLM 5.2 with no API key, no install, and no token bill—so you spend your time on real work, not pricing tables. Start here, and only graduate to a paid route if you actually outgrow it.
• Coding in an IDE/CLI all day? A GLM Coding Plan subscription is usually cheapest—just verify the standard (not promo) monthly rate.
• Building a product on the API? OpenRouter has the lowest metered rate; move to Z.AI's direct API for first-party SLAs or caching control.
• Privacy or compliance constraints? Self-host the MIT-licensed weights and pay only for hardware.

Frequently Asked Questions

How much does GLM 5.2 cost per token? About $1.40 per 1M input tokens and $4.40 per 1M output on Z.AI's official API, with cached input around $0.26. Third-party hosts like OpenRouter list less (~$0.95 / ~$3.00).

Is GLM 5.2 cheaper than Claude or GPT-5.5? Yes—on output tokens it's roughly five to six times cheaper than Claude Opus 4.8 or GPT-5.5, while staying competitive on coding benchmarks.

Is there a free way to use GLM 5.2? Three: self-host the MIT-licensed open weights, use new-user/daily credits where offered, or chat with it free in your browser at glm5.app.

What is the GLM Coding Plan? A flat monthly subscription from Z.AI (Lite through Team tiers) aimed at developers who code in an IDE or CLI all day instead of paying per token. Verify current prices on Z.AI's plan page.

Why is GLM 5.2 so much cheaper than closed models? It ships as open weights under an MIT license, which creates price competition across hosts and lets you self-host—pressure that closed, API-only models don't face.

The Bottom Line

GLM 5.2 pricing is refreshingly founder-friendly: ~$1.40/$4.40 per million tokens direct, cheaper still through OpenRouter, flat-fee Coding Plans for daily drivers, and genuinely free routes if you self-host or just open a browser. For frontier-class coding at a fifth of the closed-model price, it's hard to beat—just check live rates, because launch promos don't last.

The smartest first move costs nothing: run a real task through GLM 5.2 and see if the output quality justifies any plan at all. Do that, no keys and no setup, right here: try GLM 5.2 free on glm5.app.

Sources

• Z.AI — GLM-5.2 official documentation
• Artificial Analysis — GLM-5.2 price & performance analysis
• VentureBeat — GLM-5.2 beats GPT-5.5 for ~1/6 the cost

Pricing reflects launch-period rates (mid-June 2026). Z.AI promos, third-party providers, and subscription allowances change frequently—and GLM 5.2 consumes plan quota faster than older models—so verify the current numbers on each vendor's page before you commit.