Coding Plan vs Token Plan vs API Aggregator in November 2026: What Each Billing Model Actually Costs
AI tooling now bills in three incompatible ways, and picking the wrong billing model is usually more expensive than picking the wrong model. Priced from the vendors' own pages as of November 4, 2026: IDE coding subscriptions (Cursor Plus $8 to Ultra $200, GitHub Copilot Pro $10, Devin Desktop Pro $20, OpenCode Go $10), Chinese token and credit plans (Alibaba Bailian CNY 39 to 499, Baidu Qianfan CNY 9.9 to 600, Tencent Hy3, Volcengine Agent Plan, Xiaomi MiMo, MiniMax, Kimi Code CNY 49 to 699, Zhipu GLM Coding CNY 118 to 1078), and API aggregators (TokenRhythm, one key for 30+ models). Includes the credit-to-token conversion math, the discount deadlines that expire this month, and a decision rule mapping each billing model to the workload it actually fits.
Three billing models now sit side by side in every AI tooling procurement conversation, and they are not interchangeable in any way that matters. A coding subscription sells an IDE workflow and meters it in weekly usage pools. A token or credit plan sells raw model access and meters it in credits whose conversion to actual tokens is published unevenly at best. An API aggregator sells key consolidation and meters it in per-token rates that are sometimes below the vendor's list price and sometimes exactly the vendor's list price. All three are quoted as a monthly number, so the only comparison that survives a real invoice is to normalize into cost per million tokens.
Coding subscriptions: you are buying a workflow, not a token
Cursor restructured its tiers on October 21, 2026 and retired the old Pro and Hobby tiers, so any price list still showing Cursor Pro $20 is stale. The current structure is Free $0, Plus $8, Premium $30, Super $100 and Ultra $200. That is not a 25x price ladder over one product: the usage multipliers differ too, at 4x, 15x and 50x against Plus. One restriction matters more than the headline price, because the $8 Plus tier only exposes Grok models and any third-party model is billed separately at the API list rate. GitHub Copilot Pro at $10 runs on AI Credits where one credit is $0.01 and the Pro tier ships $10 base plus $5 flex for $15 of allowance while code completion stays unlimited. Devin Desktop Pro at $20, what Windsurf became after Cognition acquired it, allows buying extra usage at API pricing. OpenCode Go at $10 quotes its allowance per model rather than as a flat token pool, with 5-hour windows releasing 20 percent of the monthly allowance and weekly windows 50 percent.
Token and credit plans: the conversion rate is the real price
Chinese vendors converged on credit-based plans, and the headline monthly price tells you almost nothing because the credit allowance differs by more than an order of magnitude between tiers. Alibaba Cloud Bailian publishes the conversion directly: Lite at CNY 39 gets 11,500 credits a month, Standard at CNY 139 gets 45,000, and Pro at CNY 499 gets 180,000. Baidu Qianfan uses a lower entry price and a wider spread: Mini at CNY 9.9 for 1,400 points, Lite at CNY 40 for 6,600, Pro at CNY 200 for 45,000 and Max at CNY 600 for 165,000, with a 50 percent off-peak multiplier between 22:00 and 08:00. Volcengine prices in AFP points, Tencent's Hy3 floors at CNY 28 for 560 points, Xiaomi's MiMo Lite is CNY 34, MiniMax charges CNY 29 and CNY 49, and Moonshot's Kimi Code ladder runs CNY 49 to 699. Zhipu's GLM Coding Plan is the outlier that proves the point, since its three tiers now cost 141, 261 and 130 percent more than before the July 31 switch.
API aggregators: the savings are real, but not where the marketing points
TokenRhythm is the aggregator this site tracks, and its November 4 catalog is worth reading line by line. The platform no longer carries the CNY 0.80 input package for Qwen3.8-Flash that older comparison tables cite, and the discount against vendor list price is much thinner than it was. Its cheapest 1M-context routes, Qwen3.8-Flash at CNY 0.80 and CNY 2.70, GLM-5.3-Flash at CNY 0.80 and CNY 2.80 and MiMo-V2.6-Flash at CNY 1.00 and CNY 2.00, are exactly the vendor list prices rather than a discount. So the honest summary of the aggregator proposition in November 2026 is not cheaper tokens. It is one key across 30 plus models, dual protocol compatibility, image generation at CNY 0.20 per image, semantic routing, CNY 68 in signup credits, no card required to start, and upstream failover without client-side rewrites, which matters because several models here have begun retirement.
One rule for choosing, normalized to tokens
Normalizing to per-million-token cost is the only comparison that survives contact with a real invoice. On the API side the floor is CNY 0.80 input for Qwen3.8-Flash and GLM-5.3-Flash, with Zhipu currently running GLM-5.3-Flash at a limited-time 50 percent discount that puts the real rate at CNY 0.40 and CNY 1.40. On the plan side, CNY 9.9 for 1,400 points is CNY 0.0071 per point, CNY 39 for 11,500 credits is CNY 0.0034 per credit, CNY 499 for 180,000 credits is CNY 0.0028 per credit and CNY 600 for 165,000 points is CNY 0.0036 per point. Alibaba's Pro tier is the cheapest credit unit by that measure, but only for a user who consumes the full 180,000 credits. The practical rule follows directly: if your load can be described in requests per day, you are in the subscription market; if it can be described in tokens per month, you are in the API market; and if you need more than one vendor behind one key, you are in the aggregator market, where you should read the per-model catalog, not the headline.
Frequently Asked Questions
Is a coding subscription cheaper than paying for API tokens?
Not in general. Coding subscriptions sell a workflow, not tokens, so they win when your load is small and interactive: a developer editing code all day typically uses a fraction of the weekly pool and never sees a bill. They lose as soon as you run batch jobs, scheduled agents or any server-side workload, because the weekly pool is finite and resets rather than accumulating. Break-even depends entirely on consumption: Alibaba's Pro tier at CNY 0.0028 per credit is the cheapest unit rate in this market, but only if all 180,000 credits are used, and it doubles if you consume half.
What is the difference between a Token Plan and a Coding Plan?
The billing unit and the acceptable usage. A coding plan meters requests or sessions for interactive IDE work, which is why Cursor Plus at $8 only exposes Grok models and why Devin Desktop caps concurrent sessions at 10. A token plan meters credits or tokens against a published monthly allowance and can be pointed at any model the vendor hosts, which is why Alibaba quotes 11,500 to 180,000 credits per month and Baidu quotes 1,400 to 165,000 points. A token plan is what a production service should run on; a coding plan is what a human should use.
Is an API aggregator actually cheaper than the vendor's own API?
For TokenRhythm as of November 4, 2026, mostly no, and the older comparison tables in this space are stale. The cheapest 1M-context routes, Qwen3.8-Flash at CNY 0.80 and CNY 2.70, GLM-5.3-Flash at CNY 0.80 and CNY 2.80 and MiMo-V2.6-Flash at CNY 1.00 and CNY 2.00, are the vendors' own list prices, and the representative DeepSeek route is carried at the vendor's peak and off-peak rates of CNY 2.00 and CNY 8.00. What the aggregator does deliver is one key across 30 plus models, dual protocol compatibility, embeddings and the Responses API, image generation at CNY 0.20 per image, semantic routing, CNY 68 in signup credits, no card required, and upstream failover without client-side rewrites. Buy it for operational convenience, not for a discount that is no longer there.
Which discount should I act on before it expires?
Volcengine's Kimi K3.1 Pro 50 percent coefficient discount expires on November 9, 2026 and Volcengine's Coding Plan 25 percent-off promotion runs through November 8, so those two are the most urgent. Tencent Cloud's TokenHub Token Plan activity gives DeepSeek V4 Pro 30 percent off and V4.1 Flash 60 percent off until November 16 at 12:00. Alibaba's Bailian Coding Plan Pro first month at CNY 39.90 against a CNY 200 list price is new-customer only and refills daily at 09:30 Beijing time. Zhipu's GLM-5.3-Flash 50 percent discount plus 50 percent off Batch calls has no published end date, which means it can stop without warning. Tencent Cloud's Double Eleven Token Plan promotion floors at CNY 28 per month and expires November 30.
Is it worth paying for a top tier like Cursor Ultra at $200 or Alibaba Pro at CNY 499?
Only if you consume the allowance. Cursor Ultra at $200 carries a 50x Plus pool, roughly 400 times the Plus allowance, and Alibaba's Pro tier at CNY 499 buys 180,000 credits a month at CNY 0.0028 per credit, the cheapest unit rate in this market but only if all of them are used. For a solo developer, a CNY 9.9 Baidu Mini plan, a CNY 39 Alibaba Lite plan, or Cursor Plus at $8 covers ordinary interactive work at a fraction of the price. Top tiers pay off when you run agents continuously or your team hits concurrency limits on the lower tiers, not when you are writing code by hand.
What is the full domestic plan ladder with prices and allowances?
Alibaba Cloud Bailian personal tiers: Lite CNY 39 for 11,500 credits, Essential CNY 79 for 25,500, Standard CNY 139 for 45,000 with Harness benefits, Pro CNY 499 for 180,000; team seats from CNY 150. Baidu Qianfan: Mini CNY 9.9 for 1,400 points, Lite CNY 40 for 6,600, Pro CNY 200 for 45,000, Max CNY 600 for 165,000, with 50 percent off-peak from 22:00 to 08:00. Volcengine Ark: Agent Plan CNY 40 for 2,000 AFP per 5 hours and 20,000 AFP per month, first two months CNY 9.9; Coding Plan Lite CNY 40 for 148 requests per 5 hours, 320 million tokens per month and 256K context. Tencent Hy3 Token Plan CNY 28 for 560 points. Xiaomi MiMo Token Plan Lite CNY 34 continuous monthly, 1M context, no 5-hour limit, 0.8x consumption between midnight and 08:00. MiniMax Starter CNY 29 and Plus CNY 49, where Plus covers roughly 600 million M3 tokens a month. Kimi Code CNY 49, 99, 199 and 699 for 1x, 4x, 20x and 60x weekly allowance. Zhipu GLM Coding Lite CNY 118, Pro CNY 538 and Max CNY 1078 for 10,000, 60,000 and 140,000 credits a week. International equivalents: Cursor Plus $8, Premium $30, Super $100, Ultra $200; GitHub Copilot Pro $10 for $15 of allowance, Pro+ $39 and Max $100; Devin Desktop Pro $20; OpenCode Go $10, Go Plus $40 and Go Teams $20 per member; Z.ai GLM Plan $18, $80 and $168.
Want more AI pricing comparisons?
Check our price changelog for daily updates on AI model pricing, new releases, and deprecations.