Zhipu GLM-5.3 was released on August 14 and is now in Volcengine Ark Coding Plan's model switcher, with the same deduction multiplier as GLM-5.2 (source). For anyone who wants GLM-5.3, there is now a "flat monthly fee" route alongside Zhipu's own subscription and API.
This is a pricing comparison for GLM-5.3 across three routes: the flat monthly fee of Volcengine Ark Coding Plan, the credit-based Zhipu GLM Coding Plan, and the per-token Zhipu API. It breaks down how much each costs, how top-up, deals, the official entry point, quota and rate limits work, and finally whether it is worth switching to Ark for it.
Pricing comparison: what GLM-5.3 costs on each of the three routes
The conclusion up front: Volcengine Ark Coding Plan has the lowest monthly entry price (¥9.9 Lite and ¥49.9 Pro for the first two months), but it is billed per request; Zhipu bills per credit or per token, which suits users who only want the GLM family or call it on demand. The three routes use completely different billing units, so comparing monthly fees directly is misleading. Align the units first:
| Route | Billing | GLM-5.3 tier and price | Quota unit |
|---|---|---|---|
| Volcengine Ark Coding Plan | Per request + deduction multiplier | Lite ¥40/mo (¥9.9 first two months), Pro ¥200/mo (¥49.9 first two months) | 1,200–6,000 requests/5h |
| Zhipu GLM Coding Plan | Per credit + credit multiplier | Lite ¥118, Pro ¥538, Max ¥1078/mo | 2,000–28,000 credits/5h |
| Zhipu official API | Per token | Follows GLM-5.2's ¥8/¥28 (no standalone price yet) | Per 1M tokens |
Volcengine Ark Coding Plan's flat monthly fee bundles GLM-5.3 with Doubao, Kimi, MiniMax and DeepSeek into one subscription, billed per request rather than per token (source). Here is each route in detail.
Volcengine Ark Coding Plan: flat monthly fee, billed per request
GLM-5.3 uses the same deduction multiplier as GLM-5.2 on Ark, in a higher tier. In practice, running 18,000 requests entirely on GLM-5.3 burns quota faster than running them on Doubao Seed 2.1 Turbo. Volcengine has not published the exact multiplier, only that it matches GLM-5.2 and falls in the "higher multiplier, faster quota burn" tier.
On the monthly fee, Lite ¥40/mo and Pro ¥200/mo are list prices. The first two months carry a roughly 2.5x discount campaign, bringing Lite to about ¥9.9 and Pro to about ¥49.9, with list pricing restored from month three. Pro's request quota is 5x Lite's.
Zhipu GLM Coding Plan: credit-based
Zhipu's own plan is billed per credit, with GLM-5.3 multipliers of Input 6.9 / Cached 1.7 / Output 24, about 50% pricier than GLM-4.7. This is GLM-5.3's "native" entry point and also bundles web search, page reading and open-repo MCP tools (source).
Personal tiers are Lite ¥118, Pro ¥538, Max ¥1078/mo, with 20% off quarterly and 30% off annually. For the same 1M output tokens, GLM-5.3 consumes 24 credits versus only 16 for GLM-4.7.
Zhipu official API: per-token billing
On the Zhipu official API, GLM-5.3 has no standalone price yet and currently follows GLM-5.2's ¥8/¥28 reference. Cached input is ¥2 and cache storage is free for a limited time (source). This suits low-frequency calls or a self-hosted backend, not a high-frequency coding workhorse.
Deduction multiplier vs credit multiplier: why "cheap" is not the same as "worth it"
The two companies' "multipliers" are fundamentally the same thing — different models consume different quotas, just in different units. Ark folds the multiplier into requests; Zhipu folds it into credits. Because GLM-5.3 has thinking enabled by default and produces high-quality output, its multiplier ranks high on both.
So don't get carried away by "GLM-5.3 for ¥9.9". Ark's ¥9.9 Lite for the first two months is genuinely cheap, but GLM-5.3's high multiplier means you won't get many runs out of the monthly quota. Zhipu's plan costs more per month, but its credit multiplier (Output 24) is transparent, so you can do the math precisely. The two "cheap" definitions differ, which is the most easily missed point in this pricing comparison.
Top-up, deals and official entry point
The three routes have completely different top-up, deals and entry points — don't mix them up. Volcengine Ark Coding Plan's entry point is the activity page (volcengine.com/activity/codingplan), with a 2.5x first-two-month discount, 5% off for friend referrals on first order, and 5% vouchers for the referrer. Zhipu's GLM Coding Plan starts from bigmodel.cn/glm-coding, with 20% off quarterly and 30% off annually. The Zhipu official API is a balance top-up billed per token, with roughly 50% off for supported text models on Batch and a discount on cache hits.
Quota, usage and rate limits
Ark rate-limits by request count, Zhipu by credits or balance, and each uses its own key. Volcengine Ark Coding Plan uses a dedicated key; the Anthropic base URL is /api/coding and the OpenAI base URL is /api/coding/v3, effective only inside coding tools (source). Zhipu's Coding Plan uses a Coding base URL (Claude Code is https://open.bigmodel.cn/api/anthropic), while the open-platform API uses a separate API key.
Quota units: Ark Lite is about 1,200 requests/5h, 9,000/week and 18,000/month; Pro about 6,000/5h, 45,000/week and 90,000/month. Zhipu Lite is 2,000 credits/5h and 10,000/week, with off-peak calls discounted to 50% of base credits.
Is it worth it: how to choose
Conclusion: pick Volcengine Ark if you want one subscription covering GLM, Doubao, Kimi, DeepSeek and more at once; pick Zhipu's own plan if you only want the GLM family plus MCP tools; pick the Zhipu official API for low-frequency or backend self-hosting. Three decision signals:
- If you are a "multi-model scheduler" who jumps between models on one subscription, Ark's ¥9.9–49.9 monthly entry is the lowest and most worthwhile.
- If you only care about GLM and also want web search, page reading and open-repo MCP, Zhipu's own GLM Coding Plan is the only native entry that bundles those tools.
- If you call it rarely or run a backend, skip the subscription and use Zhipu API per token, with GLM-5.3 referenced at ¥8/¥28 — pay for what you use.
For a broader pricing comparison, DeepSeek API has no official monthly plan, while Kimi and Tianyi Cloud each have fixed monthly plans. GLM-5.3's value is "new flagship without a standalone price hike", and Ark's value is "the lowest entry point that bundles it with multiple models".
Caveats
Finally, three pitfalls to keep in mind — don't order just because of the ¥9.9. Specifically:
- GLM-5.3 has no standalone price on the Zhipu official API yet; ¥8/¥28 is a reference carried over from GLM-5.2, so the conclusion may change once pricing is announced.
- Volcengine Ark Coding Plan does not publish the exact GLM-5.3 deduction multiplier — only that it matches GLM-5.2 and sits in a higher tier; actual quota burn should be confirmed in the console.
- Ark Coding Plan quota only works inside coding tools; API scripts, backend and non-interactive batch calls are not allowed. Enterprise needs should use Ark API pay-as-you-go billing.
- The first-two-month 2.5x discount is a campaign; list pricing ¥40/¥200 resumes from month three, so long-term math should use list prices.