Model Studio cheatsheet
Lookup only. Learn on the map and tutorial. Concepts in the glossary.
Last verified: 2026-08-19. Trust the official page you have open for IDs, prices, and hosts.
Which path
| Job | Use | Official page |
|---|---|---|
| Call models from your service | PAYG API, OpenAI-compatible | First API call |
| Monthly quota inside coding tools | Token Plan (recommended for new buys) or Coding Plan | Token Plan, Coding Plan |
| No-code Q&A / agent | Console app, prefer Agent 2.0 | Agent 2.0 |
| Local agent, multimodal tools | CLI bl | Install |
| Chat only | Qwen (#83) | Not Model Studio |
| Alibaba coding IDE | Qoder CN / Lingma (#84) | Attach page |
Regions
China-site intro (2026-08-19): China (Beijing), US (Virginia), Singapore, Germany (Frankfurt), Japan (Tokyo).
| Fact | Source |
|---|---|
| Keys, endpoints, models, features, prices are per region | Intro |
| China-site new-user quota is Beijing only | Free quota |
| International intro puts free quota in Singapore | What is Model Studio |
| Token Plan is Beijing only | Token Plan |
| International sidebar also lists China (Hong Kong) | International intro; absent from the China-site intro. Use the site you are on |
You cannot turn the service off after activation. Stop calls by deleting keys. FAQ
API keys
| Item | Official value |
|---|---|
| Env var | DASHSCOPE_API_KEY |
| PAYG after the security upgrade | sk-ws…, shown once |
| PAYG from before the upgrade | sk-…, still valid |
| Coding Plan / Token Plan dedicated | sk-sp- |
| Beijing / Singapore / Tokyo / Frankfurt | 50 keys per primary account per region |
| US (Virginia) | 20 per owning account; no disable/reset |
| Expiry | None; delete to revoke |
| RAM user deleted | That user's keys die |
Source: Get an API key.
Base URL
Replace {WorkspaceId} with the workspace ID from the console.
| Region | OpenAI-compatible (model list / first call) | Anthropic-compatible (model list) |
|---|---|---|
| China (Beijing) | https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/compatible-mode/v1 | https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/apps/anthropic |
| Singapore | https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1 | https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/apps/anthropic |
| Japan (Tokyo) | https://{WorkspaceId}.ap-northeast-1.maas.aliyuncs.com/compatible-mode/v1 | https://{WorkspaceId}.ap-northeast-1.maas.aliyuncs.com/apps/anthropic |
| Germany (Frankfurt) | https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/compatible-mode/v1 | https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/apps/anthropic |
| US (Virginia) | https://dashscope-us.aliyuncs.com/compatible-mode/v1 (model list) | https://dashscope-us.aliyuncs.com/apps/anthropic |
DashScope native hosts use /api/v1 on the same model-list page.
Official pages disagree — this table picks a winner:
- The intro sample still shows
{WorkspaceId}.us-east-1.maas.aliyuncs.comfor Virginia. The US row follows the model list. - Coding Plan still lists PAYG OpenAI as
https://dashscope.aliyuncs.com/compatible-mode/v1. New code follows first call / model list.
Coding Plan dedicated:
- OpenAI:
https://coding.dashscope.aliyuncs.com/v1 - Anthropic:
https://coding.dashscope.aliyuncs.com/apps/anthropic
Token Plan dedicated hosts come from the Token Plan console. Do not reuse the Coding Plan row.
How to look up models
Do not maintain a full catalog here.
Intro tiers: Max = strongest; Plus = default recommendation; Flash = low latency. The 2026-08 intro names qwen3.8-max. The first-call sample uses qwen-plus.
FAQ: qwen-plus-latest is Qwen3, not an alias of Qwen3.5 / Qwen3.7. Those are sibling series.
Plan comparison (numbers move; checkout wins)
From Token Plan and Coding Plan as opened 2026-08-19:
| Token Plan Personal | Token Plan Team | Coding Plan Pro | |
|---|---|---|---|
| Meter | Credits, 7-day window | Credits / seat / month | Request counts (5-hour / week / month caps; first hit wins) |
| List price on that day | Lite CNY 39/mo promo (60); Standard 139 (180); Pro 499 (600) | Standard seat 150 promo (198); Pro 550 (698); Max 1,398 | CNY 200/mo; the page has mentioned a 39.90 first month — checkout wins |
| Region | Beijing only | Beijing only | See that page |
| Training data | Personal is used to improve the service (intro "Important") | Team: no training-use commitment | Inputs and outputs are used to improve the service |
| Refunds | That page / the agreement | That page / the agreement | Official: no refunds |
| Status | Official recommendation for new buys | Multi-seat | Lite: no new buys after 2026-03-20, no renewals after 2026-04-13; Pro is limited stock |
Coding Plan Pro caps on that page: 6,000 / 5 hours, 45,000 / week, 90,000 / month. Simple tasks ~5–10 requests; hard ones ~10–30+.
PAYG realtime offset order: new-user quota > resource pack > savings plan > account balance. Free quota, Billing
Batch is 50% of realtime and cannot use free quota / savings / packs.
Wanxiang membership does not offset Model Studio API. FAQ
Model Studio CLI
| Item | Official value |
|---|---|
| npm package | bailian-cli |
| Commands | bl, bailian |
| Node | ≥ 22.12.0 |
| Install | npm install -g bailian-cli (not pnpm / yarn for this package) |
| Login | bl auth login --console |
| Smoke test | bl text chat --message "ping" --non-interactive --output json |
| Region flag | --region cn|us|intl, default cn |
Not aliyun bailian.
Glossary index
One line each; definitions on the glossary page.
| Term | Hook |
|---|---|
| Model Studio | Platform, not a chat window |
| Qwen | Model family and another product |
| Lingma / Qoder CN | IDE |
| DashScope | Legacy API brand; the env var remains |
| Workspace | Isolation for ACL and bills |
| Three protocols | OpenAI / Anthropic / DashScope |
| Token Plan vs Coding Plan | Two subscriptions that cannot migrate |
| Agent 1.0 vs 2.0 | No upgrade path |
| Two CLIs | bl ≠ aliyun bailian |
Sources
Last systematic pass: 2026-08-19.
S: official source of truth
| Source | Use |
|---|---|
| What is Model Studio (ZH) | Definition, regions, billing boundary |
| What is Model Studio (EN) | International nav and regions |
| Models | IDs, regions, three base URLs |
| First API call | Activation, samples |
| Get an API key | Key format, caps, env var |
| Billing items | Inference / train / deploy |
| Pricing | Unit prices |
| Free quota | 90 days, stop switch |
| Token Plan | Credits plans |
| Coding Plan | Request plans, dedicated URL |
| Clients | Third-party tools |
| CLI install | bailian-cli |
| FAQ | Activate / cannot disable / vs Qwen |
A: official but marketing or slow
| Source | Use |
|---|---|
| China product page | Positioning; numbers lose to the help center |
| Console | Real activation and quota |
| CLI landing | CLI capabilities |
| Privacy | Cross-check plan data-use text |
| Service agreement | §5.2 (named by Coding Plan) |
B: verify
Community posts. IDs and hosts rot fast. Commands come from S only.