Couldn’t Get the GLM Coding Plan? Volcengine’s ¥9.9 Offer Can Drop to About ¥9.4 with an Invitation

Short answer

If the GLM Coding Plan is unavailable when you need it, refreshing the purchase page is not your only option. Volcengine Ark currently advertises the Coding Plan Lite subscription at ¥9.9 per month for the first two months. The same page lists a regular monthly price of ¥40, making the introductory price roughly 25% of the regular price, or “2.5 zhe” in Chinese discount terminology. The plan page also lists GLM-5.2 and says selected popular models receive a temporary four-times quota boost.

Read the fine print before buying: ¥9.9 is not a permanent price, the third month returns to the displayed ¥40 monthly rate, and Coding Plan quota is intended for supported coding tools rather than arbitrary general-purpose API traffic. This article and its screenshots were prepared on July 11, 2026. The official campaign page displayed an end date of August 7, 2026; always trust the live checkout page if the rules change.

Why this small promotion deserves a full article

AI coding subscriptions increasingly resemble mobile data plans. Developers no longer want to calculate the cost of every individual token. They want a predictable monthly budget that lets a coding agent inspect repositories, edit files, call tools, run tests, read failures, and try again.

The problem is familiar to anyone who has tried to buy a popular discounted plan: availability can be limited, campaign entries move, and the cheapest tier may be difficult to obtain at exactly the moment a real project needs it. Most developers are not emotionally attached to a billing page. They need four practical outcomes: a low trial cost, a model capable of real code work, compatibility with an existing agent, and a migration path that does not require rebuilding the entire workstation.

Volcengine’s current offer is interesting because it combines a low two-month entry price with a multi-model pool. Think of it as a food court card rather than a coupon for one restaurant. The official page currently lists GLM-5.2 alongside Kimi, Doubao, DeepSeek, and MiniMax models. That does not make every model identical, but it gives developers more room to compare them inside one supported ecosystem.

What does ¥9.9 actually buy?

On July 11, 2026, the official campaign page showed a regular Lite monthly subscription price of ¥40 and an introductory price of ¥9.9 for the first two months. Since 9.9 divided by 40 equals 24.75%, the commonly advertised “2.5 zhe” means the subscription costs about one quarter of the displayed regular price during the introductory period.

Four numbers describe four different things:

• ¥9.9 is the monthly charge for the first two months.
• ¥40 is the regular monthly price displayed for later renewals.
• “2.5 zhe” describes the subscription discount, not a separate GLM-5.2 model discount.
• “Four-times quota” describes a temporary usage multiplier for selected popular models.

A simple analogy helps. A coffee shop may sell the first two drinks for ¥9.9 and upgrade them to an extra-large cup. The cheap receipt and the larger cup are two separate benefits. The third drink does not automatically stay cheap, and the larger cup does not mean the cashier applies another discount.

How the effective price reaches about ¥9.4

¥9.9 is the public campaign price, but an invited user making an eligible first subscription can receive an additional 5% discount. The arithmetic is:

¥9.9 × 95% = ¥9.405

That is why the campaign is described as “as low as approximately ¥9.4.” Depending on checkout rounding, the payment page may show ¥9.40 or ¥9.41. Eligibility, stacking, account restrictions, and the final amount must always be confirmed on the live checkout page.

To use the invitation:

• Open the invitation image or the text link below.
• Sign in to Volcengine and enter the Coding Plan campaign page.
• Confirm that invitation code NZ38JNT2 is recognized.
• Select the Lite recurring monthly subscription.
• Verify that the additional 5% invitation discount appears before payment.

Text purchase link: Subscribe to Volcengine Coding Plan with invitation code NZ38JNT2

Referral disclosure: This is an invitation link. Under the rules currently shown on the campaign page, an eligible invited user receives an additional 5% discount on the first subscription, while the inviter may receive a coupon equal to 5% of the friend’s order amount. You can still use the ordinary campaign entrance if you prefer.

Think of the public campaign as reducing a ¥40 product to ¥9.9, followed by a separate 5%-off checkout coupon. The invitation discount applies to the already reduced ¥9.9 amount, not directly to the original ¥40 price.

Why GLM-5.2 is the important part

The strongest part of this campaign is not simply the number of models. It is that GLM-5.2 appears directly in the supported model list. Someone originally looking for a GLM Coding Plan can therefore separate two questions: “Can I use GLM-5.2 for coding?” and “Can I purchase one specific vendor’s subscription today?”

Coding-agent workloads are much heavier than ordinary chat. A request such as “fix this failing feature” may require directory discovery, reading a dozen files, tracing function calls, editing code, running tests, studying the error, and repeating the loop. It is closer to asking a courier to move boxes around a warehouse than asking someone to deliver one envelope. Quota multipliers matter because they determine how many complete work loops fit into a month.

The lower price does not remove security obligations. Do not expose production credentials, customer data, private certificates, or unredacted configuration files to an agent. Use ignored files, narrow permissions, test repositories, and sanitized copies. A subscription solves a budget problem; it does not automatically solve data governance.

Is it really a replacement for the GLM Coding Plan?

It is better described as an alternative purchase path, not an identical product.

A developer who prefers Zhipu’s native account, console, support path, and product integration may still prefer the original GLM plan when it is available. A developer whose actual requirement is inexpensive GLM-5.2 access inside a coding agent may find Volcengine more flexible, especially when comparing multiple models.

Imagine two transit products. One is a monthly pass issued by a single bus operator: simple, direct, and consistent. The other is a city transit card accepted by several operators: broader choice, but with different routing and accounting rules. Both can reach the destination, yet their quotas, supported tools, billing systems, and support boundaries are not the same.

Use this decision order:

• If native GLM integration matters most and the original plan is available, compare real quotas and supported tools before choosing.
• If the original plan is unavailable and project work cannot wait, the Volcengine offer is a low-cost fallback.
• If you frequently compare models, a multi-model pool may be more valuable than a single-model subscription.
• If you only need a general application API, verify the product boundary before purchasing a Coding Plan.

Four traps that matter more than the headline

Treating the introductory price as permanent

A recurring monthly subscription can continue charging after the campaign period. Create two calendar reminders immediately after purchase: one seven days before the second month ends and another one day before renewal. Even if you intend to keep the plan, reassess model quality, quota, and real usage before the regular price begins.

Treating Coding Plan quota as general API credit

The official FAQ explains that plan resources are for supported coding tools and are not equivalent to unrestricted platform API balance. It is like a gym membership that works in selected facilities, not a bank card accepted in every shop. A production application should use the appropriate metered API or product resource instead of assuming the coding subscription covers it.

Mixing compatible protocols

Many tools support both OpenAI-compatible and Anthropic-compatible endpoints. The model may be the same, but environment variables and base URLs are not interchangeable. Think of one building with a front entrance and an underground garage: both lead inside, but each follows different access rules. The scripts below use Anthropic-style variables commonly used by Claude Code.

Giving the full API key to an agent

An agent can prepare configuration files, but the secret should be entered locally through a hidden terminal prompt, environment variable, or operating-system credential store. Never paste a complete key into chat, screenshots, Git, articles, or group messages. The supplied scripts do not upload secrets or call a third-party configuration service, but the resulting shell profile still needs restrictive file permissions.

One-click configuration for three operating systems

These scripts configure the common Anthropic-compatible environment variables used by Claude Code and select glm-5.2 as the default model. They do not upload configuration, scan the local network, or depend on an external configuration service. You still need an official Coding Plan API key and an installed coding agent.

• Windows 11 PowerShell script
• Ubuntu 26.04 Bash script
• macOS 26 Zsh script

Windows 11:

powershell -ExecutionPolicy Bypass -File .\setup-volcengine-coding-plan-windows11.ps1

Ubuntu 26.04:

chmod +x setup-volcengine-coding-plan-ubuntu2604.sh
./setup-volcengine-coding-plan-ubuntu2604.sh
source ~/.profile

macOS 26:

chmod +x setup-volcengine-coding-plan-macos26.zsh
./setup-volcengine-coding-plan-macos26.zsh
source ~/.zshrc

The Linux and macOS scripts back up or replace their own marked configuration block. The Windows script writes user-level environment variables. If the official console later changes a model identifier or endpoint, use the live “Quick Start” values rather than copying an old article blindly.

Manual automation: visible and reversible

If you prefer not to run a downloaded script, perform the same sequence manually:

• Open the Coding Plan console and copy the current Anthropic-compatible base URL.
• Create or copy a dedicated Coding Plan API key instead of reusing another project’s credential.
• Set ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN, and ANTHROPIC_MODEL in the current user environment.
• Restart the terminal so the new environment is loaded.
• Ask the agent to create, read, and delete a temporary text file in a temporary directory.
• Confirm that usage appears in the console and that GLM-5.2 handled the request.
• Only then allow the agent to enter a real repository.

This is the same logic used when installing a new lock. Test it with an empty box before placing valuables behind the door.

Agent-driven configuration without surrendering the secret

Give the following prompt to a local coding agent. Let it prepare files and backups, but pause when the API key must be entered:

Inspect the current operating system and default shell. Configure the current user
for Volcengine Ark Coding Plan through its Anthropic-compatible endpoint.
Requirements:
1. Back up every file before changing it.
2. Use the base URL currently shown in the official console.
3. Set glm-5.2 as the default model.
4. Never print the complete API key in chat, logs, or shell history.
5. Pause and let me enter the key through a hidden terminal prompt.
6. Test only in a temporary directory, not a real repository.
7. Report changed files and exact rollback commands.

This is like hiring a technician to install a safe while the owner privately enters the final combination. The agent handles operating-system differences without needing to learn the secret.

Is the plan worth buying?

For an existing coding-agent user, two Lite months at ¥9.9 each provide a cheap experiment. The important test is not whether GLM-5.2 can answer a clever prompt. Test whether it can understand your repository, make controlled edits, run your tests, recover from failures, and stay within a useful monthly quota.

Cheap is not the same as valuable. If you ask two simple questions per month, free options may already be enough. If an agent repeatedly reads a repository and executes long repair loops, predictable monthly pricing becomes much more meaningful.

My practical recommendation is to buy only if you can complete setup, a small benchmark, quota observation, and a renewal reminder on the same day. Do not collect AI subscriptions like unused gym memberships.

Q&A

Does the ¥9.9 price continue forever?

No. At the time of writing, the official page displayed ¥9.9 per month for the first two months and ¥40 per month afterward. Campaign details can change, so verify the checkout page.

Why can ¥9.9 become approximately ¥9.4?

An eligible first subscription opened through the invitation receives an additional 5% discount. Since 9.9 × 0.95 equals 9.405, the marketing figure is approximately ¥9.4. The checkout page remains the source of truth.

Is GLM-5.2 itself discounted to 25%?

Not exactly. The 25% figure comes from the plan price, ¥9.9 compared with ¥40. The separate GLM-5.2 promotion is described as a temporary quota multiplier. Price and quota are different benefits.

Can I use Claude Code?

The campaign page lists Claude Code and other supported coding tools and provides compatible connection methods. The exact tool list can change; use the live Quick Start page.

Can my application use this key as normal API credit?

Do not assume so. The official FAQ limits Coding Plan resources to supported coding tools. Production integrations should use the appropriate platform API billing arrangement.

Is GLM-5.2 always the best coding model?

No model wins every repository and task. Front-end generation, long-context understanding, tool calls, Chinese requirements, and test repair can produce different rankings. Use the same sanitized benchmark across models.

How do I avoid an unexpected third-month charge?

Create calendar reminders on the purchase date and review the payment platform’s renewal controls. Whether you disable renewal immediately depends on the platform rules and your intended use, but do not wait for the invoice to remember.

References

• Official Volcengine Ark Coding Plan campaign: https://www.volcengine.com/activity/codingplan
• Invitation purchase entry, code NZ38JNT2: https://volcengine.com/L/L6-GBhQcLPQ/
• Official help and Quick Start links: accessible from the campaign page’s Quick Start, discount rules, quota rules, and plan details entries
• Screenshot and price verification date: July 11, 2026

Campaign prices disappear, models change, and quotas evolve. The durable lesson is the process: verify the billing period, verify the usage boundary, test the smallest possible task, and only then connect the plan to a real project.