Qwen3.8-Max (GA)
GAAlibaba’s 2.4T next-generation flagship, now generally available (the preview was retired; old model IDs auto-route). 1M context; officially described as capable of autonomous coding for days to deliver complete projects.
Back to comparisonQwen Token
Alibaba Cloud · Qwen
Qwen Token
Plan Guide
The request-based Coding Plan has been fully upgraded to a Credits-based Token Plan: Qwen3.8-Max (2.4T, 1M context) is now generally available, joined by qwen3.8-flash, deepseek-v4-pro-0813 and image/audio models sharing one Credits pool. Personal plans start at ¥39/mo early-bird; night usage (22:00-08:00) is half price.
¥39Early-bird entry
40,000Pro Credits/week
18Models
8Agents (Pro)
Models
Qwen3.8-Max now GA, plus qwen3.8-flash, deepseek-v4-pro-0813 and the Qwen / DeepSeek / GLM / image / audio families.
qwen3.8-flash (new)
NewThe new low-cost high-speed tier of the Qwen3.8 family.
deepseek-v4-pro-0813
NewDeepSeek-V4-Pro 0813 version snapshot, newly added.
qwen3.7-max / qwen3.7-plus
Qwen3.7 reasoning and general-purpose flagships.
qwen-image-3.0-pro / qwen-audio-3.0
Night 50% offNew image-generation and real-time audio/ASR models, plus wan2.7 image and happyhorse-1.1 video, all sharing the same Credits.
GLM family
Zhipu flagship models (glm-5.2).
Personal plans
Early-bird prices (Lite 35% / Standard 23% / Pro 17% off) with Credits on a 7-day rolling window; team seats also available.
Lite
Early bird ¥39/mo¥39/mo
List ¥60/mo (35% off)
- 2,500 Credits per 7-day window
- 1-2 concurrent Agents
- All multimodal models share the quota
- Night 50% off
Best for
Subscribe to Lite →Light users and first-timers
Standard
Recommended¥139/mo
List ¥180/mo (23% off) · 4x Lite
- 10,000 Credits per 7-day window
- 3-4 concurrent Agents
- All Lite benefits
- Night 50% off
Best for
Subscribe to Standard →Daily development workhorse
Pro
Heavy duty¥499/mo
List ¥600/mo (17% off) · 16x Lite
- 40,000 Credits per 7-day window
- 6-8 concurrent Agents
- All Standard benefits
- Night 50% off
Best for
Subscribe to Pro →Heavy and multi-Agent workloads
Tier comparison
| Tier | Early-bird | List | Credits/7d | Agents |
|---|---|---|---|---|
| Lite | ¥39 | ¥60 | 2,500 | 1-2 |
| Standard | ¥139 | ¥180 | 10,000 | 3-4 |
| Pro | ¥499 | ¥600 | 40,000 | 6-8 |
| Top-up pack | ¥100/pack | — | 20,000/pack | Unlimited |
Quota and access rules
Early-bird prices are limited-time. Calls to qwen3.8-max and deepseek-v4-pro-0813 between 22:00 and 08:00 deduct Credits at 50%. A dedicated sk-sp- API Key and Base URL are required; generic Bailian keys do not deduct plan Credits.
7-day rolling window
A 7-day timer starts at your first call; hitting the cap pauses service until the window resets.
Night 50% off
Every night 22:00-08:00, qwen3.8-max and deepseek-v4-pro-0813 deduct half Credits — schedule heavy jobs accordingly.
No training
In-plan calls are not used for model training.
What happened to the old Coding Plan
- The request-count based Coding Plan (Lite ¥40 / Pro ¥200) was upgraded to the Credits-based Token Plan
- The old Lite discontinuation and daily 9:30 restock model no longer applies
- Existing users migrate per official guidance; historical quotas follow the order page
Frequently Asked Questions
How is Token Plan different from the old Coding Plan?
Alibaba Cloud upgraded the request-count based Coding Plan into a Credits-based Token Plan: models expanded from Qwen3.5 to 18 models led by the now-GA Qwen3.8-Max, plus qwen3.8-flash, deepseek-v4-pro-0813 and multimodal families, with quotas switched to model-rate Credits and personal/team versions.
Which personal tier should I pick?
Lite ¥39/mo (2,500 Credits/7d) for light use, Standard ¥139/mo (10,000 Credits/7d, 4x Lite) as the daily workhorse, Pro ¥499/mo (40,000 Credits/7d, 16x Lite) for heavy multi-Agent workloads. All are limited-time early-bird prices.
How are Credits consumed?
Each model has its own Credits rate, accumulated over a 7-day rolling window starting from your first call. Night usage (22:00-08:00) of qwen3.8-max and deepseek-v4-pro-0813 deducts 50% fewer Credits.
Is there a team version?
Yes — team seats: Standard ¥150/seat/mo (25,000 Credits/mo), Advanced ¥550 (100,000) and Premium ¥1,398 (250,000, early-bird), with a 29-model pool for centralized company billing.
What happened to the 9:30 AM restock?
It no longer applies. The limited-restock model of the old Coding Plan Pro was retired with the upgrade; Token Plan can be subscribed directly.
Is there a refund policy?
Activated plans are generally non-refundable. Confirm the rules before purchase.