ChinaModelAPI

News / Model Watch · GLM commercialization

Monetization milestone GLM Coding Plan ¥118/月起 Tmall AI 空间站 Sep 3 首个大模型天猫旗舰店
2026-09-04 · Model Watch · 新浪财经 / 凤凰网科技 / 金融界 / 电商派 multi-source, Sep 2-3

Zhipu Opens a Tmall Flagship: GLM Coding Plan From ¥118/Month — Tokens Enter the Age of "Phone Credit" Retail

On September 2, Zhipu became the first Chinese LLM vendor with an official Tmall flagship store (store license of record: Beijing Zhipu), selling GLM Coding Plan subscriptions and token top-ups — personal tiers from ¥118/month, team tiers from ¥598. A day later Tmall itself launched 「AI空间站」, a token top-up center with Alibaba Cloud and Zhipu as first vendors. Buying model compute now works like topping up a phone. The read for the API market: distribution mechanics just changed, while model pricing itself didn't.

Direct answer

Zhipu's Tmall flagship (Sep 2) sells GLM Coding Plan tiers — Lite/Pro/Max on monthly, quarterly and yearly cycles — from ¥118/month (personal) and ¥598 (team), plus token top-ups. Chinese financial coverage flagged the Pro tier at ¥538/month ≈ 3.6x its original direct-sale price as a pricing controversy. Tmall's follow-up AI 空间站 (Sep 3) turns token retail into a platform category — first vendors: Alibaba Cloud, Zhipu. No change to GLM model availability or API rates; this is a channel and brand play on the road Zhipu has been walking all quarter (MaaS volume, coding-plan subscriptions, GLM-5.3 open weights).

What's confirmed across sources

  • The store: live since Sep 2 on Tmall; the store's business registration lists 北京智谱 as operator — official flagship, not a reseller.
  • What it sells: GLM Coding Plan subscriptions in Lite / Pro / Max tiers, monthly / quarterly / yearly billing; token top-up packages. Searchable in Taobao as 「智谱旗舰店」.
  • Pricing: personal from ¥118/month; team from ¥598. 金融界 reported the Pro tier at ¥538/month ≈ 3.6x the original price — framed by media as "selling tokens is more like advertising" (新浪财经) given the mark-up discussion.
  • Tmall's move: AI 空间站 (token top-up center) launched Sep 3, Alibaba Cloud + Zhipu first; promo-season discounts on tokens become mechanically possible.
  • Pattern context: 36kr pairs it with "Kimi 找微软" — Chinese labs are visibly professionalizing monetization this month: Zhipu goes retail, Moonshot negotiates hosting-revenue shares (K3 30% ask).

Why it matters — and to whom

  • Tokens become giftable, discountable, procurable. E-commerce rails bring discovery browsing, corporate purchase orders, 618/Double-11 promos — mechanics API portals never had. Expect the other labs to follow onto 空间站.
  • The Pro-tier mark-up is the tell. Retail pricing above direct pricing means the store optimizes for reach and gifting, not undercutting the API — consistent with the "advertising" read. Devs comparing costs should still benchmark the direct rate card.
  • International users are untouched. Tmall requires Chinese payment rails; GLM-5.3 international access stays on z.ai USD pricing ($1.40/$4.40 per 1M) or OpenAI-compatible relays with global billing.
  • For relay operators: a consumer-grade on-ramp grows the GLM user base — eventual spillover demand for OpenAI-compatible, globally-billed access is the second-order effect worth tracking.

Primary sources

FAQ (2026)

What launched Sep 2?

Zhipu's official Tmall flagship — GLM Coding Plan (Lite/Pro/Max, month/quarter/year) and token top-ups. First LLM vendor on Tmall.

Pricing?

Personal from ¥118/month, team from ¥598; Pro tier ¥538/month ≈ 3.6x original direct price (controversy per 金融界).

What's AI 空间站?

Tmall's token top-up center (Sep 3) — buy LLM subscriptions like phone credit; Alibaba Cloud + Zhipu first vendors.

Why it matters?

Tokens inherit e-commerce mechanics — gifting, promos, procurement. Media reads the store as brand play as much as revenue.

For international devs?

No practical access (CN payment rails). GLM-5.3 stays on z.ai USD rates ($1.40/$4.40) or OpenAI-compatible relays with global billing.

Model-side changes?

None — GLM-5.3 / 5.3-Flash APIs, rates and open weights are unchanged. This is a new channel, not a new model.

Related guides