Zhipu’s Tmall Store Turns AI Coding Credits Into a Retail Subscription

Zhipu AI’s Tmall flagship store is important less because it changes what a coding model can do than because it changes how AI capacity is sold. Instead of acquiring access only through a developer-oriented platform, customers can browse a visible plan, add it to an e-commerce cart, and pay through a mainstream retail channel. Reporting described it as the first instance of a listed large-model company placing its core AI product on a mainstream e-commerce shelf. 1114

What Zhipu is selling on Tmall

The flagship store launched on September 2 with GLM Coding Plan subscriptions: Lite, Pro, Max, and a standard team offering. Reported individual list prices are ¥118, ¥538, and ¥1,078 per month; the standard team plan starts at ¥598 per seat per month. Customers can choose monthly, quarterly, or annual billing.

This is not simply a prepaid packet of raw API tokens. The Coding Plan is a subscription with credit-based limits. Reports describe concurrent rolling five-hour and weekly allowance windows. For example, the reported Lite allocation is 2,000 credits per five hours and 10,000 per week; Pro is 12,000 and 60,000; Max is 28,000 and 140,000.

Some coverage inconsistently labels the 10,000-credit Lite allocation as monthly rather than weekly. The tier comparison and plan-focused reports describe it as a weekly limit, so shoppers should confirm the current storefront terms before purchasing.

Why credit-based plans matter to developers

A credit system makes the trade-off between price and usage more explicit than a loosely defined “unlimited” plan. Developers can select a capacity tier based on expected workload and must account for both short-window and weekly limits. Reported plan details also indicate that off-peak use can consume credits at half the standard rate. 18

The product is designed around coding-agent workflows rather than a single Zhipu interface. Coverage of the Tmall listing says the GLM-5.3-based plans work with more than 20 agent tools, including ZCode, Claude Code, and Codex. Separately, plan documentation notes that Coding Plan access is distinct from SDK-based or custom API integrations and may be restricted to supported tools. 17

That distinction matters: Zhipu is packaging a developer workflow subscription, while its broader platform also sells model usage through API and Model-as-a-Service offerings.

From constrained access to a retail channel

The Tmall launch follows a period in which coding-model demand and capacity management were central to Zhipu’s commercial story. Industry reports describe the Coding Plan as beginning in September 2025, encountering rationing amid GLM-4.7-related demand in January 2026, and returning to open sales on July 31. Those product-history details are based on industry reporting rather than the company’s financial disclosures, so they should be treated cautiously. 7

Even so, the business logic of the store is clear: a retail marketplace can turn a specialist purchase into an always-available, price-segmented subscription. It reduces friction for individual developers and smaller teams while giving Zhipu a prominent distribution channel beyond its own developer surfaces.

Tmall moved quickly to build out the category. The day after Zhipu’s launch, it introduced an AI Space Station, or Token Recharge Center, offering subscription products from multiple model providers. 11

What the launch says about Zhipu’s business

The storefront arrives as Zhipu scales its cloud-model business following its January Hong Kong listing. The company reported first-half 2026 revenue of ¥953.9 million, up 400% year over year, according to Reuters. 1 Reporting on its interim results says open-platform and API revenue accounted for 86.5% of first-half revenue, evidence of a sharp move toward cloud-delivered model usage. 4

But rapid revenue growth should not be confused with demonstrated profitability. Zhipu reported a first-half net loss of roughly ¥2.0 billion, even as the loss narrowed from the prior year. 14

Tmall should therefore be viewed as a distribution and customer-acquisition experiment within a rapidly growing usage business—not as proof that retail subscriptions solve the economics of frontier-model development. The important test will be whether retail buyers renew, increase their usage, and do so at margins that support the infrastructure and research costs behind the service.

A sign of China’s emerging token-commerce market

The broader significance is cultural as well as commercial. AI usage is increasingly presented as a recurring, metered service that can be purchased in consumer-style plans—closer to topping up mobile data or a video membership than negotiating an enterprise software contract.

That framing is emerging against a large backdrop of model consumption. The National Data Administration was reported to have said China processed about 140 trillion tokens daily in March 2026, up from 100 billion at the start of 2024. 15 The figure does not show that Tmall sales will become a major revenue source for Zhipu, but it helps explain why providers and platforms are experimenting with simpler ways to package and sell compute.

For developers, the practical takeaway is straightforward: compare the plan’s supported tools, rolling credit limits, billing term, and API eligibility—not just the headline price. For the market, Zhipu’s storefront is a concrete signal that AI model access is becoming a retail product category as well as an enterprise and API service.