Find Your Perfect AI Coding Plan
Coding Plan = Client Tool + API Subscription + Dev Environment. One setup, full productivity.
Why Coding Plans Matter
The right plan elevates your coding ability, cost efficiency, and experience
Slash Costs
China-native Coding Plans are often cheaper than subscribing to overseas IDEs directly, and fit local payment and network conditions better
Switch Models Freely
One subscription unlocks Doubao, DeepSeek, Qwen, Kimi and more — switch anytime
China-Native Access
Ark and Bailian are China-native — no VPN needed, stable low-latency access
How to Build a Plan
Three steps to start AI-assisted coding
Pick a Client
Choose your preferred IDE plugin or CLI tool (Cline, Cursor, Claude Code…)
Connect an API
Subscribe to a Coding Plan, get your API key, paste into client settings
Start Coding
Enjoy AI-assisted coding, switch models freely, upgrade plans as needed
🔥 Recommended Coding / Agent Plan APIs
Mainstream China-native AI coding & Agent subscriptions covering coding, agents, multimodal and multi-seat use cases. Pricing and promos vary, check official sites.
火山方舟 Coding Plan
🔥 Official promoOn 2026-11-03 the discount-rules page https://www.volcengine.com/docs/82379/2525065 (page UpdatedTime 2026-07-27) and the activity page https://www.volcengine.com/L/s3lNTNYxaEc/ were opened. The rules table lists Lite at 40 yuan list / 9.9 yuan promo and Pro at 200 yuan list / 49.9 yuan promo. The window is 2026-06-08 through 2026-11-08. Each account gets at most two promo months, shared across new purchase, renewal, and upgrade, then list price. The referral promo can stack an extra 5% off (the page says 9.5折), and it does not print the combined amount. The old 8.9 yuan figure and the three-month half-price claim were not on these pages. The activity page's monthly Lite card shows 9.9 against 40/month. The Pro monthly card's priceConfig also says 9.9 and 40/month, while its SKU is Coding_Plan_Pro_monthly. That conflicts with the rules table and with the quarterly 600 and annual 2400 strikethrough prices. This file keeps Pro at 200 from the rules table. Quarterly and annual strikethrough prices equal the monthly list times 3 or 12 (Lite 120/quarter and 480/year; Pro 600/quarter and 2400/year). The monthly tab image filename says 2.5折, matching the rules-page text. The quarterly and annual tab image filenames say 4.9折 and 8.7折, but the card amounts are 116.4 and 582 per quarter and 456 and 2280 per year, which do not match those two rates, so 4.9折 and 8.7折 are not recorded as current discount rates. The roughly 1,200 / 6,000 requests per 5 hours were not on the docs or the activity page and remain the earlier record. The plan overview names a 5-hour limit, a weekly limit resetting Monday 00:00:00, and a monthly limit resetting at 00:00:00 on day 1 of the subscription month. It does not give request counts or a timezone. current_version is still the old label v2. These pages do not print a version string.
Volcano Ark Agent Plan
Agent BundleStandard monthly prices are ¥40/200/500/1000 for Small/Medium/Large/Max. AFP is not tokens or request count. Small/Medium exclude video; multimodal/Harness daily quotas are half the monthly quota, subject to the model table. Exhaustion blocks use unless overage postpaid is enabled, which can switch automatically and incur extra charges. Model multipliers/promotions vary; old blanket 7.5× and companion-gift claims are not retained. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.
MiniMax Token Plan (Legacy)
Legacy Token PlanThis entry reviews Token Plan, which is closed to new sales, not M Plan as the same current product. There is no universally purchasable new-user Token Plan quote; existing charges and extras depend on the renewal contract. The official legacy notice confirms M3.1 Flash and H3 Image, but not H3 video. Continuous renewal retains legacy prices and benefits, with no legacy tier changes; once disabled, renewal cannot restart. Exception: iOS App Store subscriptions cannot renew after the current term; vendor compensation credits are valid for one year. Migration to M Plan is irreversible and forfeits legacy extras. Plus can move to Go/Explore/Build, Max to Explore/Build, Ultra to Build, with unused term value credited. New M Plan amounts were not verified: this is NOT a full pricing review of that new product, and old Starter ¥29 offers are not current prices. Verify retention, training and account quotas under the contract; there is no zero-processing promise or automatic equivalence to M Plan quotas. Historical scores are not a new-sale recommendation. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.
百炼 Coding Plan
💜 Lite no longer soldLegacy Coding Plan Pro is normally ¥200/mo. Docs currently advertise a limited first month at ¥39.90 for new subscribers, subject to checkout, not a recurring price. Quotas of 6,000/5h, 45,000/week and 90,000/month count model calls, not user prompts. Lite new purchases/renewals have stopped. Requires dedicated sk-sp- credentials and coding.dashscope endpoints, separate from standard Bailian credentials. Only permitted interactive AI-tool use; no scripts, app backends or noninteractive batch API calls. No refunds/account sharing; inputs and outputs may be used for service/model improvement. Authorized RAM subaccounts can now subscribe. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.
Bailian Token Plan (Team)
👥 Team / Multi-seatTeam Standard/Premium/Enterprise seats are now ¥150/550/1298 per month for 25,000/100,000/250,000 Credits. Each seat is one member with a dedicated key; keys and seat quotas cannot be shared. A shared pack is ¥5,000/625,000 Credits, valid one month. Credits are not tokens. Exhausting all plan/shared credits blocks calls without automatic pay-as-you-go; unused credits do not roll over. Beijing region only; individual/team plans bill independently. Historical introductory/multi-seat discounts were not confirmed in current entitlement docs and are removed. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.
Zhipu GLM Coding Plan
💎 GLM FlagshipMonthly prices for new personal credit plans. On 2026-11-03 the subscribe page showed both ¥94.4/¥118, ¥430.4/¥538, and ¥862.4/¥1078, each labeled per month. 94.4, 430.4, and 862.4 match the previously recorded quarterly 20%-off equivalents, but this pass did not see the words for a quarterly plan or for 20% off. The annual 30%-off equivalents (¥82.6/¥376.6/¥754.6) were not reprinted and remain the 2026-09-28 record. The usage page says the 5-hour credits refresh 5 hours after consumption, and weekly credits refresh every 7 days from the order. Peak hours are Monday–Friday 14:00–18:00 UTC+8, and off-peak model calls cost 50% of base credits. Deduction coefficients per 10,000 tokens are Input/Cached/Output 6.9/1.7/24 for GLM-5.3 and 2.3/0.56/8 for GLM-5.3-Flash. The tools page marks ZCode and AutoClaw at 1.5× quota. OpenClaw and other general agents are secondary scheduling and can be queued or rate-limited under load. Legacy V1/V2 pricing and renewal eligibility follow the official notice. Use is limited to supported tools; checkout prices and temporary promotions follow the official site.
DeepSeek API (Pay-as-you-go)
⚡ Pay-as-you-goDocumentary review 2026-11-18: this is a USD usage-based API, not a fixed monthly or unlimited plan. Current pricing still lists separate deepseek-flash (V4.1-Flash-1017) and deepseek-v4-pro (V4-Pro-0813); the older September news post’s Pro-rerouting statement conflicts with current pricing and the updated changelog. Current documentation takes precedence; backend routing was not tested. Off-peak is half of peak. Peak hours are Monday–Friday 09:00–12:00 and 14:00–18:00 Asia/Shanghai; weekends and Chinese public holidays are off-peak all day. Search and Fetch each cost $0.75 per 1,000 successful requests producing results, including calls through Harness, Claude Code or Codex (including Alpha Search), with possible additional model-token charges. The former free period ending 2026-11-01 00:00 Asia/Shanghai has expired; do not budget these calls as free. File parsing/OCR costs $0.225/1,000 pages. TTS lists $0.45/10,000 characters, temporarily $0.225; ASR lists $0.116/hour, temporarily $0.058. No end date for those temporary prices is stated. Legacy Flash IDs route to current Flash; deepseek-chat / deepseek-reasoner are not listed as current models and their availability is unknown. Listed integrations are the official guides checked here, not an exhaustive or runtime-tested list. The published general privacy policy includes improvement/training purposes, a right to opt out of training, and retention varying by purpose and legal requirements, not an API-specific zero-retention guarantee. Account-specific contracts and regional connectivity were not verified. Historical editorial scores retained without new hands-on testing, not a new performance or privacy certification.
30-Second Picker
Answer 3 questions to find your best AI coding setup
1. Your dev environment?
📋 All Coding Plans
16 plans, sorted by overall score