DashScope API Endpoints: Coding Plan Base URL, Key Format & Routing Rules (2026)
DashScope API endpoints split into two incompatible paths: standard dashscope.aliyuncs.com (sk- key) and Coding Plan coding.dashscope.aliyuncs.com (sk-sp- key).
Archive item produced with AI assistance from the cited source and published without individual review. Editor of record: Joe Werner.

Most teams plugging a new API key into a gateway assume the key just works — same endpoint, same billing dashboard, same quota counters. DashScope's Coding Plan breaks that assumption deliberately, and if your routing layer doesn't account for it, you will quietly accrue pay-per-use charges while burning a subscription your quota dashboard never sees.
What happened
Alibaba Cloud has consolidated its AI model subscription offering for coding tools under Coding Plan Pro (¥200/month). The plan aggregates access to Qwen3.6-plus, Kimi K2.5, GLM-5, MiniMax-M2.5, and several other models under a flat-rate, request-capped structure aimed at coding-agent workflows — Claude Code, OpenClaw, Cursor, and similar tools.
The critical infrastructure detail: Coding Plan issues a dedicated API key and base URL that are entirely separate from the standard DashScope pay-per-use endpoint.
| Attribute | Standard DashScope | Coding Plan |
|---|---|---|
| API key prefix | sk- | sk-sp- |
| OpenAI base URL | https://dashscope.aliyuncs.com/compatible-mode/v1 | https://coding.dashscope.aliyuncs.com/v1 |
| Anthropic base URL | (not applicable at standard endpoint) | https://coding.dashscope.aliyuncs.com/apps/anthropic |
| Quota tracking | Per-token billing | Request-count against plan quota |
| API call automation | Allowed | Prohibited |
Alibaba explicitly warns: using the wrong key/URL combination causes the standard endpoint to treat the request as a regular pay-per-use call. The Coding Plan key will not work at the standard URL, and the standard key will not draw from subscription quota.
The plan's quota structure uses rolling windows: 6,000 requests per 5 hours (rolling recovery), 45,000 per week, 90,000 per month. Each "request" may internally expand to 5–30+ model calls depending on task complexity, context size, and tool use.
The Lite tier has been discontinued: new purchases stopped March 20, 2026; renewals ended April 13, 2026.
Why it matters for AI engineering teams
1. Gateway pass-through is a billing trap. If a team routes sk-sp-* keys through a standard OpenAI-compatible gateway — including any router that reformulates headers or rewrites base URLs — the subscription quota is not consumed and the request either fails or falls through to a pay-per-use path. DashScope's warning is explicit: mixing keys and base URLs is the single most common misconfiguration and results in unexpected charges.
2. The automation prohibition is real. The Coding Plan terms prohibit using the subscription API key in automated scripts, custom application backends, or any non-interactive batch scenario. Violation can suspend the subscription or ban the key. This means the plan is architecturally unsuitable for server-side routing through a gateway — it is designed for human-paced, IDE-resident coding agent traffic.
3. Multi-region complicates key management further. DashScope has four regional endpoints (Beijing, Virginia, Singapore, Frankfurt). The standard endpoints are region-specific; API keys are not cross-region. Coding Plan endpoints appear to be Beijing-only. Teams operating globally need a clear separation between which models are routed through standard DashScope regional endpoints (via a gateway) and which are accessed directly by coding tools.
4. Data consent differs. Coding Plan terms include a clause authorizing Alibaba Cloud to use input and output content for model training and service improvement. This differs from the standard DashScope agreement. Teams with data governance requirements need to evaluate this separately before enrolling developer machines in Coding Plan.
5. RAM sub-account provisioning. Coding Plan must be purchased by the main account; RAM sub-accounts require explicit permissions. This creates an admin dependency — developer-provisioned coding agents cannot self-serve subscription access.
The router/operator angle
The Coding Plan architecture draws a sharp line between two routing regimes:
Gateway-routed traffic (standard DashScope, pay-per-use): qwen3.7-max, qwen3.6-plus, qwen3.6-flash, and third-party models like deepseek-v4-pro and kimi-k2.6 accessed at dashscope.aliyuncs.com/compatible-mode/v1. This is the path where a router adds value — fallback, latency hedging, provider cost comparison, and unified billing.
Coding Plan subscription traffic: accessed directly from the coding tool at coding.dashscope.aliyuncs.com. No gateway should sit between the tool and this endpoint. The subscription's request-count quota tracking only works with the dedicated key/URL pair, and the terms prohibit gateway automation.
For teams using TheRouter to centralize DashScope traffic, the implication is direct: do not route Coding Plan keys through TheRouter. Use TheRouter for DashScope pay-per-use API calls and reserve Coding Plan direct configuration for the developer workstation tools that Alibaba designed it for.
The practical separation for an engineering team:
- Production pipelines, evals, batch jobs, agent backends: route through gateway with standard
sk-key anddashscope.aliyuncs.comendpoint - Developer IDE coding agents (Claude Code, Cursor, OpenClaw): configure directly with
sk-sp-key andcoding.dashscope.aliyuncs.comendpoint — never through a proxy
A clean routing policy documents this boundary explicitly, so new developers don't accidentally enter the wrong key in the wrong field.
What TheRouter users should watch or try
If you're routing to Qwen models through TheRouter today using standard DashScope credentials, your setup is unaffected by Coding Plan. Continue using the standard DashScope provider config with your sk- API key.
If your team is evaluating Coding Plan for developer coding tools:
- Verify key format before configuration. The
sk-sp-prefix is the fastest indicator you have a Coding Plan key — do not route it through any gateway. - Audit your routing configs for base URL. If any config file points to
coding.dashscope.aliyuncs.comand routes through a proxy layer, remove the proxy hop. - Review the data consent clause at the Coding Plan subscription page before enrolling if your team has data governance requirements or contracts that restrict training data use.
- Watch for request-count burn rate. The 6,000 requests/5h window can exhaust quickly with complex coding-agent workflows. Monitor usage on the Coding Plan dashboard to tune session frequency.
- Plan for Lite sunset fallout. If any developer still holds a Lite plan key (expired April 13), those keys are inactive. Audit your internal Cursor/Claude Code config distribution to confirm no Lite keys remain deployed.
Quick Reference: DashScope API Endpoints
What are the DashScope API endpoints?
DashScope exposes two incompatible endpoint families:
| Use case | Base URL | Key prefix |
|---|---|---|
| Standard OpenAI-compatible API | https://dashscope.aliyuncs.com/compatible-mode/v1 | sk- |
| Coding Plan OpenAI-compatible API | https://coding.dashscope.aliyuncs.com/v1 | sk-sp- |
| Coding Plan Anthropic-compatible API | https://coding.dashscope.aliyuncs.com/apps/anthropic | sk-sp- |
What is the difference between DashScope and Qwen API?
DashScope is Alibaba Cloud's model-serving platform; Qwen is the model family hosted on it. When you call the DashScope API, you access Qwen (and third-party) models through DashScope's billing and routing infrastructure. The two terms are not interchangeable at the API level — DashScope provides the endpoint, Qwen is one of the models you call through it.
Which DashScope endpoint should a coding agent use?
If you are using Coding Plan (the flat-rate subscription for coding tools), configure your agent to call coding.dashscope.aliyuncs.com with the sk-sp- key directly — do not route through a gateway. All other DashScope traffic (pipeline, eval, batch) should use the standard dashscope.aliyuncs.com endpoint through your router.

DashScope Managed Agents API Goes Commercial: What the New Workspace-Scoped Agent Runtime Means for Your Routing Architecture
Alibaba Cloud's Managed Agents API enters commercial billing on August 17. The workspace-scoped agent runtime introduces a new endpoint namespace and SDK gate every DashScope operator must audit now.

wan2.7 T2V Video Generation on DashScope: Async Job Routing and Region Deployment Guide
Alibaba Cloud DashScope's wan2.7-t2v model supports text-, image-, and reference-to-video generation via async job APIs across mainland China, international, and US scopes. Pick your region, job polling strategy, and fallback paths before production.

OpenAI Agents API Beta: The New Gateway Bypass Operators Need to Account For
OpenAI's Agents API beta ships a dedicated session harness under client.beta.agents that doesn't route through /v1/chat/completions. For teams using AI gateways, this means billing blind spots, missing audit trails, and a new API key scope to manage.