You can buy tokens from Tianyi Cloud two ways today: TeleAI Token Plan subscriptions, with 11 tiers across three lines starting at ¥9.90/mo; or Xirang API Tokens packages, five models sold as fixed token packs also from ¥9.9 (10M tokens/1 month). The pricing logic of the two systems is completely different—subscriptions buy a monthly quota pool plus product capabilities, while packages buy fixed token amounts that only count down after activation—so you cannot compare the numbers directly. First decide whether you work inside coding tools or need backend API calls. All prices below are as of the official pages on 2026-08-18.
Tianyi Cloud Token Plan Pricing Comparison: Subscription and Xirang API Are Two Systems
You cannot look up one number for Tianyi Cloud Token Plan pricing—the coding Token Plan, Token Plan packages and Xirang API Tokens packages are separate billing systems, and subscription quota does not equal package balance. The coding Token Plan uses dedicated keys and monthly quota for coding tools only; Token Plan packages and Lite/Standard/Premium target API and business calls; Xirang API packages offset online inference and must never be confused with coding-plan keys (source). High-level comparison first:
| Dimension | Tianyi Cloud Token Plan Subscription | Xirang API (Tokens Packages) |
|---|---|---|
| Billing model | Monthly subscription, fixed fee for calendar-month quota | Fixed token packages, per-pack billing |
| Official entry | Tianyi Cloud AI campaign page | Token Service docs / console |
| Purchase | Buy plan in console, manual renewal | Buy package in console, manual activation |
| Quota | Shared pool offsets input and output | Package quota offsets both input and output tokens |
| Entry price | ¥9.90/unit/mo (Lite, 10M tokens) | ¥9.9 per 10M tokens/1 month |
| Models | GLM-5 (formal), DeepSeek-V3.2 (flagship) | DeepSeek-V4-Flash, GLM-5, DeepSeek-V3.2, DeepSeek-V3, DeepSeek-R1 |
| Best for | Coding tools or API business calls | Backend online inference |
Subscriptions follow a "quota pool" logic: a fixed monthly fee buys a calendar-month allowance, and one main account holds one plan at a time. Packages follow a "pre-purchase" logic: fixed token amounts that must be manually activated before validity starts counting. The clearest difference: subscriptions reset monthly at a fixed price; packages are buy-what-you-use with validity starting on activation. Our Tianyi Cloud page aggregates all tiers and rules.
TeleAI Token Plan Membership Pricing: Choosing Among 11 Tiers in Three Lines
Tianyi Cloud Token Plan now spans three lines and 11 tiers: five coding Token Plan tiers (¥29–699/mo) with GLM-5 and DeepSeek-V3.2, three Token Plan packages (GLM5, ¥39.90–299.90 per unit/mo) with 200K context, and three Lite/Standard/Premium tiers (DeepSeek v3.2 quota packs, ¥9.90–49.90 per unit/mo); the former GLM Coding Plan is sold out. Tier pricing (source):
| Tier | Price | Monthly quota | Best for |
|---|---|---|---|
| Coding Token Plan 25M | ¥29/mo | 25M tokens | Entry experience, AI coding beginners |
| Coding Token Plan 80M | ¥89/mo | 80M tokens | Daily development, script maintenance |
| Coding Token Plan 180M | ¥199/mo | 180M tokens | Main tier: complex projects, sustained iteration (recommended) |
| Coding Token Plan 380M | ¥399/mo | 380M tokens | Professional development, project teams |
| Coding Token Plan 680M | ¥699/mo | 680M tokens | Heavy development, parallel projects |
| Line | Tier | Price | Monthly quota | Notes |
|---|---|---|---|---|
| Token Plan packages (GLM5) | Basic | ¥39.90/unit/mo | 15M tokens | 200K context, agent building |
| Token Plan packages (GLM5) | Pro | ¥159.90/unit/mo | 70M tokens | 200K context, engineering for teams |
| Token Plan packages (GLM5) | Flagship | ¥299.90/unit/mo | 150M tokens | 200K context, enterprise-scale docs |
| Lite / Standard / Premium | Lite | ¥9.90/unit/mo | 10M tokens | DeepSeek v3.2, personal trial |
| Lite / Standard / Premium | Standard | ¥29.90/unit/mo | 40M tokens | DeepSeek v3.2, daily API |
| Lite / Standard / Premium | Premium | ¥49.90/unit/mo | 80M tokens | DeepSeek v3.2, batch processing |
How to pick: for coding tools, the 180M tier (¥199/mo) is the most balanced long-term choice, and the 25M tier (¥29/mo) or Lite pack (¥9.90) is a fine low-cost trial; for agent building, knowledge-base Q&A and long context, look at the Token Plan packages (GLM5); for daily API and batch calls, use Lite/Standard/Premium. There are no quarterly or annual discount tiers on the official site; one main account holds one plan at a time, plans take effect immediately, renewal is manual during validity (no auto-renewal yet), repurchase after expiry, and no refunds except as required by law.
How Much Does Xirang API Cost: Tokens Package Pricing Across Five Models
Xirang API bills through Tokens packages—all five models (DeepSeek-V4-Flash, GLM-5, DeepSeek-V3.2, DeepSeek-V3, DeepSeek-R1) are sold as fixed token packs from ¥9.9 (10M tokens/1 month), with quota covering both input and output tokens. Full price table (source):
| Model | Package tiers | Validity | List price |
|---|---|---|---|
| DeepSeek-V4-Flash | 10M / 32M / 100M tokens | 1 month | ¥9.9 / ¥29.9 / ¥89.9 |
| GLM-5 (formal) | 15M / 70M / 150M tokens | 1 month | ¥39.9 / ¥159.9 / ¥299.9 |
| DeepSeek-V3.2 (flagship) | 10M / 40M / 80M tokens | 1 month | ¥9.9 / ¥29.9 / ¥49.9 |
| DeepSeek-V3 | 10M / 50M / 100M / 500M / 1B tokens | 1/3/6/12 months | ¥35 / ¥175 / ¥350 / ¥1750 / ¥3500 |
| DeepSeek-R1 | 10M / 50M / 100M / 500M / 1B tokens | 1/3/6/12 months | ¥70 / ¥350 / ¥700 / ¥3500 / ¥7000 |
Three takeaways: first, DeepSeek-V3.2 (flagship) is the official recommendation—10M tokens at ¥9.9, same entry as V4-Flash but positioned as value-driven code generation; second, DeepSeek-V3 and DeepSeek-R1 offer 1/3/6/12-month validity bands for steady batch workloads; third, each account can hold at most 2 packages per model, and packages must be manually activated—validity starts on activation. Tool capabilities include OpenAI-compatible API, Chat, Embeddings, Reranker, Batch inference and context cache.
Top-up, Quota and Rate Limits: Two Different Accounting Models
Subscriptions are about "monthly quota pool + manual renewal," while packages are about "fixed quota + manual activation + 2-pack-per-model cap"—two completely different consumption and reset models. The coding Token Plan provides calendar-month quota, takes effect immediately, renews manually during validity, and requires repurchase after expiry, with one plan per main account; Token Plan packages and Lite/Standard/Premium are monthly subscriptions too. Xirang API packages do not auto-activate—activate them in the console, after which status turns "in use" and validity starts; with multiple packs, "pending" ones can be deactivated (source).
Do not compare only entry prices: for a 10M-token package, DeepSeek-V4-Flash and DeepSeek-V3.2 cost ¥9.9, GLM-5's corresponding tier is ¥39.9 (15M tier), DeepSeek-V3 is ¥35 and DeepSeek-R1 is ¥70—model positioning, not just pack size, moves the price by an order of magnitude. Coding plans instead bundle GLM-5 and DeepSeek-V3.2 into one shared quota pool, so pick by scenario rather than by unit price.
API Key and Integration Config: Coding Toolchain vs Open API
The coding Token Plan uses dedicated keys for Claude Code, Cursor, Cline, TRAE, CodeBuddy and other coding tools against monthly quota; Xirang API uses OpenAI-compatible integration with regular API keys for Chat, Embeddings, Reranker, Batch and context cache, aimed at self-built backends. The two key systems and billing are not interchangeable. This is the easiest pitfall: backends must use Xirang API keys and package credits—reusing a coding-plan key will break billing (source).
Setup splits by scenario: coding-tool users fill in the dedicated key and model (GLM-5 / DeepSeek-V3.2) in their tool; backend developers enable the online inference service in the console, buy and activate a package, then configure base URL and model name in OpenAI-compatible format. Token Plan packages (GLM5) emphasize 200K context for agent building and knowledge-base Q&A; Lite/Standard/Premium emphasize DeepSeek v3.2 fixed low-rate deduction, with the official claim of over 60% lower cost than pay-as-you-go for batch calls.
Deals and Is It Worth It: How to Buy Tianyi Cloud AI Membership Smartly
Bottom line: for coding, go with the coding Token Plan (180M tier at ¥199/mo recommended); for API and business calls, go with Xirang API packages (DeepSeek-V3.2 from ¥9.9)—pick by scenario and never mix the two. Current deals are mainly the 500K free-tokens entry displayed on the campaign page and "limited-time deal" labels on the coding Token Plan tiers; there are no quarterly or annual discounted rates. The former GLM Coding Plan (Lite/Pro/Max, ¥49/149/469 monthly) is sold out—do not buy the old plan from third-party channels (source).
| Your situation | Recommendation |
|---|---|
| First-time AI coding, low budget | Coding Token Plan 25M (¥29/mo) or Lite pack (¥9.90/unit/mo) |
| Daily development mainline | Coding Token Plan 180M (¥199/mo, dual models) |
| Team heavy development | Coding Token Plan 680M (¥699/mo) |
| Agent building, knowledge-base Q&A, long context | Token Plan packages (GLM5, 200K context, from ¥39.90/unit/mo) |
| Backend API calls, batch inference | Xirang API packages (DeepSeek-V3.2 from ¥9.9; deep reasoning on DeepSeek-R1 from ¥70) |
For broader comparisons against other Chinese platforms, check the DeepSeek, Zhipu GLM, Qwen and Kimi pricing pages. Whether Tianyi Cloud Token Plan is worth it depends on which path you take: a subscription trades a monthly budget for a fixed quota pool, while packages pre-buy fixed tokens—separate your coding scenario from your API scenario first, then compare prices.