FAQ | Qwen Cloud

Important resources

Coding Plan page: Manage your subscription and view usage on the Coding Plan page
API Keys page: View and reset your API keys on the API Keys page
Supported models: See the full list of models
Setup guides: Configuration instructions for Claude Code, OpenCode, OpenClaw, and more

Start here: 60-second diagnosis

If your tool is not working and you see errors like 401, 403, 404, invalid access token, invalid api-key, quota exceeded, or connection failures, check these first:

API key: Must start with sk-sp-, not sk-
Base URL: Must contain coding
Tool type: Coding Plan works only in interactive coding tools
Quota: Check whether your 5-hour, weekly, or monthly quota is exhausted
Updated config: Restart the tool after updating the API key or base URL

If you are still blocked, continue with the questions below.

Most common questions

Why am I still billed after subscribing?

If you still see pay-as-you-go charges after subscribing, your tool is likely using the wrong API key, base URL, or cached credentials. Check in order:

Your API key starts with sk-sp-
Your base URL uses coding-intl.dashscope.aliyuncs.com
Your tool is not falling back to old general credentials
You restarted the tool after changing the configuration

Use the Coding Plan page to verify usage. If needed, reset your key on the API Keys page.

Which API key and base URL should I use?

If you are unsure about endpoint setup, key format, or whether dashscope-intl can be reused, use this rule: Coding Plan requires a plan-specific key plus a coding endpoint. Use a plan-specific key and a coding base URL:

API key: sk-sp-xxxxx
OpenAI-compatible: https://coding-intl.dashscope.aliyuncs.com/v1
Anthropic-compatible: https://coding-intl.dashscope.aliyuncs.com/apps/anthropic

General dashscope-intl URLs do not work with Coding Plan keys.

I got 401 or 403. Is my API key invalid?

If you see 401 invalid access token, HTTP 401: Incorrect API key, or 403 invalid api-key, the problem is usually a wrong key type, an expired subscription, a malformed key, or a key and URL mismatch. Check these points:

Confirm the key starts with sk-sp-
Re-copy the key to remove hidden spaces or line breaks
Confirm your subscription is active
Make sure the base URL contains coding

In most cases:

401 invalid access token means wrong key type, expired subscription, or malformed key
403 invalid api-key means you are using a Coding Plan key with a non-Coding Plan URL

Why do I get 404 or connection errors?

If you get 404 status code (no body), endpoint not found, or connection errors, the base URL path is usually wrong for your tool. Use the correct path for your protocol:

Anthropic-compatible tools use /apps/anthropic or /apps/anthropic/v1
OpenAI-compatible tools use /v1

Also check that the domain is exactly coding-intl.dashscope.aliyuncs.com.

Why does the tool say quota exceeded?

If you see hour allocated quota exceeded, week allocated quota exceeded, or month allocated quota exceeded, your request quota has been exhausted. Coding Plan uses request-based limits:

6,000 requests per 5 hours
45,000 requests per week
90,000 requests per month

Reset times:

5-hour quota: 5 hours after exhaustion
Weekly quota: Monday 00:00:00 (UTC+8)
Monthly quota: same day next month at 00:00:00 (UTC+8)

There is no automatic fallback to pay-as-you-go when quota is exhausted.

Can I use Coding Plan via cURL, Postman, Dify?

No. Coding Plan is for interactive coding tools such as Claude Code, OpenCode, OpenClaw, Cursor, Cline, and Qwen Code. It is not supported for generic API testing tools or server-side scripts.

Common errors & fixes

How to handle input length limits?

If you get 400 InvalidParameter: Range of input length should be [1, xxx], your context or request exceeds model limits. Try these fixes:

Create a new session
Switch to a model with a longer context window
In OpenCode, configure limit to restrict context length

See Overview for supported models.

Why is my model "not supported"?

If you get model 'xxx' is not supported, the model ID is usually incorrect, case-sensitive, or copied with extra spaces. Check these points:

Use the exact model name from Overview
Remove any leading or trailing spaces
Verify capitalization exactly as documented

How do I enable thinking mode?

If you want reasoning or thinking mode, setup differs by tool.

Claude Code
OpenCode
Qwen Code
OpenClaw

Type /config, go to Thinking mode, and set it to true.To view the thinking process, press Ctrl + O.

In your model configuration, add thinking inside the model's options field:

{
  "provider": {
    "qwen-cloud-coding-plan": {
      "models": {
        "qwen3.5-plus": {
          "options": {
            "thinking": {
              "type": "enabled",
              "budgetTokens": 8192
            }
          }
        }
      }
    }
  }
}

The thinking configuration must be nested under provider → models → <model-id> → options. For the complete configuration structure, see the OpenCode setup guide.

Use version v2026.03.02 or later
Type /think high
Add reasoning and compat to the model definition in models.providers.qwen-cloud.models:

{
  "models": {
    "providers": {
      "qwen-cloud": {
        "models": [
          {
            "id": "qwen3.5-plus",
            "name": "qwen3.5-plus",
            "reasoning": false,
            "compat": {
              "thinkingFormat": "qwen"
            }
          }
        ]
      }
    }
  }
}

The reasoning and compat fields go inside each model object in the models array, not under agents.defaults.model.

Restart OpenClaw:

openclaw gateway restart

For the complete configuration, see the OpenClaw setup guide.

Why is my ModelScope token rejected?

If you see Authentication failed, please make sure that a valid ModelScope token is supplied., your tool is connected to ModelScope instead of Coding Plan. ModelScope tokens (ms-xxx) and Qwen Cloud keys (sk-xxx or sk-sp-xxx) are not interchangeable. For Coding Plan, follow your tool's setup guide. For ModelScope, get tokens from modelscope.ai.

Why is my plan restricted to Coding Agents?

If you see Coding Plan is currently only available for Coding Agents, your request source is unsupported. Coding Plan is for interactive coding tools only. It is not supported in curl, Postman, Dify, or custom scripts.

Claude Code

Why am I prompted to migrate from npm?

If you see Claude Code has switched from npm to native installer., run:

claude install

Then follow the prompts.

Why is Claude Code connection failing?

If startup fails with Unable to connect to Anthropic services or ERR_BAD_REQUEST, Claude onboarding likely did not complete in your environment. Claude Code tries to connect to api.anthropic.com on first startup, which fails in regions where Claude is unavailable. Fix it like this:

Add "hasCompletedOnboarding": true as a top-level field in ~/.claude.json:

{
  "hasCompletedOnboarding": true
}

You can also use Qwen Code to add this automatically. Run qwen and send: Please add "hasCompletedOnboarding": true as a top-level field in ~/.claude.json without overwriting existing content.

Save the file
Restart Claude Code with claude

OpenCode

How to resolve "Request Entity Too Large"?

If you see Request Entity Too Large, your request payload or context is too large. Try these fixes:

Run /new for a new conversation
Run /compact to compress context
Update OpenCode to version 1.2.16 or later

How to fix "Thinking budget" errors?

If you get The thinking_budget parameter must be a positive integer and not greater than 81920., your budgetTokens value is too large for the selected model.

Model	Max thinking budget
qwen3.5-plus	81,920
qwen3-max-2026-01-23	81,920
kimi-k2.5	81,920
glm-5	32,768
glm-4.7	32,768
MiniMax-M2.5	32,768 (chain-of-thought + response)
qwen3-coder-next	Not supported
qwen3-coder-plus	Not supported

Update your model configuration in opencode.json so options.thinking.budgetTokens stays within the limit. For unsupported models, remove the options.thinking configuration.

How do I fix max_tokens exceeded in OpenCode?

If you get InternalError.Algo.InvalidParameter: Range of max_tokens should be [1, xxxx]., your max_tokens or limit setting is too large. Adjust the limit value in opencode.json. For more details, see OpenCode.

Cursor

Why can't I use Coding Plan models in Cursor?

If you see:

The model xxx does not work with your current plan or API key
Named models unavailable. Free plans can only use Auto. Switch to Auto or upgrade plans to continue

Cursor Free only supports Auto mode. Upgrade to Cursor Pro or higher for custom models.

Qoder

Why can't I find the Model option in Qoder?

OpenClaw

Why is OpenClaw calling unexpected models?

If OpenClaw is making unexpected model calls, model restrictions are usually missing from the default configuration. Declare the allowed models in ~/.openclaw/openclaw.json. See OpenClaw for the expected setup.

Why does OpenClaw connect but hang?

If OpenClaw shows Connected but keeps spinning or never returns a response, the issue is usually a local proxy or stale cached model config. Common causes:

A local proxy is intercepting requests to coding-intl.dashscope.aliyuncs.com
An incorrect base URL is cached in ~/.openclaw/agents/main/agent/models.json

Try this:

Disable the proxy or make sure it can access the domain
Delete the cache file and restart OpenClaw

rm ~/.openclaw/agents/main/agent/models.json
openclaw gateway restart

How to fix OAuth token refresh failures?

If you see Agent failed before reply: OAuth token refresh failed., OpenClaw is using a third-party OAuth service instead of Coding Plan. Fix it like this:

Set up Coding Plan in OpenClaw using the setup guide
Clear old credentials

rm ~/.openclaw/agents/main/agent/auth.json
rm ~/.openclaw/agents/main/agent/models.json
openclaw gateway restart

Why does OpenClaw say Unknown model?

If you see Agent failed before reply: Unknown model., the provider key name, model prefix, or old config is usually wrong. Check in order:

models.providers includes a qwen-cloud key
agents.defaults.model.primary starts with qwen-cloud/
Old providers like qwen-portal have been removed

Example:

{
  "models": { "providers": { "qwen-cloud": {...} } },
  "agents": { "defaults": { "model": { "primary": "qwen-cloud/qwen3.5-plus" } } }
}

How to handle API rate limits?

If you see API rate limit reached., the issue can be a configuration mismatch, exhausted quota, or an API key problem. Check in order:

Verify that the base URL and provider setup match the OpenClaw guide
Check usage on the Coding Plan page
Reset your API key on the API Keys page if needed

Why does model discovery fail in OpenClaw?

If you see Failed to discover Alibaba Cloud models: TimeoutError or 404 Not Found, this does not affect normal use. To hide the message, remove alibaba-cloud:default from ~/.openclaw/agents/main/agent/auth-profiles.json so it looks like this:

"profiles": {}

How do I fix auth.profiles errors in OpenClaw?

If you see errors like:

auth.profiles.qwen-portal:default.mode: Invalid input
auth.profiles.qwen-portal:default: Unrecognized key: "apiKey"

Remove the incorrect auth.profiles entries. API keys belong in models.providers.qwen-cloud.apiKey, not in auth.

How do I view models configured for Coding Plan in OpenClaw?

Run openclaw tui, then type /model to list and select models.

Why do I get "HTTP 401: Incorrect API key provided." in OpenClaw?

Possible causes:

Your API key is invalid, expired, empty, incorrectly formatted, or mismatched with the endpoint environment. Confirm it is the Coding Plan-specific key. Copy it fully without spaces. Confirm your subscription is active.
Stale cache. Delete providers.qwen-cloud from ~/.openclaw/agents/main/agent/models.json and restart OpenClaw.

How do I add Coding Plan models without losing OpenClaw config?

Do not replace the entire file — this overwrites custom settings. Make partial edits instead. Two methods:

OpenClaw works: Enter the instruction below in an OpenClaw chat.
OpenClaw has no model or doesn't respond: Install Qwen Code, start it, and enter the instruction below. Instruction (replace YOUR_API_KEY with your actual API key):

Configure OpenClaw to connect to Coding Plan. Merge the following into ~/.openclaw/openclaw.json (keep existing settings. Add new fields if missing):
1. Under models.providers, add "qwen-cloud":
   - baseUrl: https://coding-intl.dashscope.aliyuncs.com/v1
   - apiKey: YOUR_API_KEY (replace with actual value)
   - api: openai-completions
   - models.mode: merge
   - Models list (each model has matching id and name, reasoning is false, cost is 0):
     - qwen3.5-plus: input ["text","image"], contextWindow 1000000, maxTokens 65536
     - qwen3-max-2026-01-23: input ["text"], contextWindow 262144, maxTokens 65536
     - qwen3-coder-next: input ["text"], contextWindow 262144, maxTokens 65536
     - qwen3-coder-plus: input ["text"], contextWindow 1000000, maxTokens 65536
     - MiniMax-M2.5: input ["text"], contextWindow 196608, maxTokens 32768
     - glm-5: input ["text"], contextWindow 202752, maxTokens 16384
     - glm-4.7: input ["text"], contextWindow 202752, maxTokens 16384
     - kimi-k2.5: input ["text","image"], contextWindow 262144, maxTokens 32768

2. Under agents.defaults, set:
   - model.primary: "qwen-cloud/qwen3.5-plus"
   - models: all models above (format: "qwen-cloud/model-id": {})

3. API Key: YOUR_API_KEY
4. baseUrl: https://coding-intl.dashscope.aliyuncs.com/v1

After configuration, run openclaw gateway restart to apply changes.

Verify with openclaw models status in a new session. Restart existing sessions after restarting the gateway.

Billing and quota

When do my quotas reset?

Coding Plan has three quota windows:

Quota Type	Limit	Reset Time
5-hour limit	6,000 requests	5 hours after exhaustion
Weekly limit	45,000 requests	Monday 00:00:00 (UTC+8)
Monthly limit	90,000 requests	Same day next month, 00:00:00 (UTC+8)

If you hit hour allocated quota exceeded, week allocated quota exceeded, or month allocated quota exceeded, wait for the corresponding reset time.

Does Coding Plan auto-switch to PAYG?

No. When quota is exhausted, API calls fail directly. There is no automatic fallback to pay-as-you-go.

When does my subscription expire?

Your subscription starts at activation and expires at 23:59:59 (UTC+8) on the same date next month. If the next month does not have that date, the subscription expires on the last day of that month. Example:

Activate on February 3
Expires on March 3

How can I view my usage?

Check the Coding Plan page for overall usage. Per-model and token-level details are not available.

Can I buy an annual plan?

No. Coding Plan is monthly only.

Is there a free trial for Coding Plan?

No. Coding Plan does not participate in Qwen Cloud's free trial program.

Can I renew early?

Yes. If you renew before expiration, the new term extends from your current expiration date. Example:

First activation: March 15 → expires April 15
Renew on April 10 → extends to May 15

If you renew after expiration, you must repurchase and your API key will change.

Can I use coupons when I renew?

Yes. General-purpose coupons are accepted for renewal.

Will I get reminders before my subscription expires?

Yes. Automatic reminders are sent 7, 3, and 1 day before expiration through in-site messages, email, SMS, and phone.

Product features

Which models are supported by Coding Plan?

Recommended models:

qwen3.5-plus (vision)
kimi-k2.5 (vision)
glm-5
MiniMax-M2.5

Additional supported models:

qwen3-max-2026-01-23
qwen3-coder-next
qwen3-coder-plus
glm-4.7

See the latest list in Overview.

What are the model context limits?

Model	Context window (tokens)
qwen3.5-plus	1,000,000
qwen3-coder-plus	1,000,000
qwen3-max-2026-01-23	262,144
qwen3-coder-next	262,144
kimi-k2.5	262,144
MiniMax-M2.5	196,608
glm-5	202,752
glm-4.7	202,752

To reduce context limit errors:

Use models with longer context windows, such as qwen3.5-plus or qwen3-coder-plus
Start tools in a specific project folder
Break large tasks into smaller subtasks
Use precise, focused instructions

How do I reset or manage my API key?

Use the Reset button on the API Keys page to reset your key. Keep these rules in mind:

After reset, update the key in all tools
If you resubscribe after expiration, the key changes
Coding Plan supports one key per subscription

Can I manage multiple API keys/IP lists?

No. Coding Plan does not support multiple keys or an IP allowlist.

Can I have multiple active subscriptions?

No. Each account can have one active Coding Plan subscription. No. Coding Plan is for personal use only. API keys cannot be shared. Public exposure or account sharing may result in automatic key suspension.

Which tools are compatible with Coding Plan?

Coding Plan works with interactive coding tools that support OpenAI-compatible or Anthropic-compatible APIs. Examples include:

Claude Code
OpenCode
OpenClaw
Cursor
Cline
Qwen Code
VS Code extensions that support OpenAI or Anthropic APIs

It is not supported in Dify, Postman, curl, or custom backend automation.

Are the models full-featured or quantized?

All supported models in Coding Plan are full-featured, unquantized versions.

What should I do if I see data_inspection_failed?

If you see data_inspection_failed, refer to Troubleshooting data inspection errors.

​Important resources

​Start here: 60-second diagnosis

​Most common questions

​Why am I still billed after subscribing?

​Which API key and base URL should I use?

​I got 401 or 403. Is my API key invalid?

​Why do I get 404 or connection errors?

​Why does the tool say quota exceeded?

​Can I use Coding Plan via cURL, Postman, Dify?

​Common errors & fixes

​How to handle input length limits?

​Why is my model "not supported"?

​How do I enable thinking mode?

​Why is my ModelScope token rejected?

​Why is my plan restricted to Coding Agents?

​Claude Code

​Why am I prompted to migrate from npm?

​Why is Claude Code connection failing?

​OpenCode

​How to resolve "Request Entity Too Large"?

​How to fix "Thinking budget" errors?

​How do I fix max_tokens exceeded in OpenCode?

​Cursor

​Why can't I use Coding Plan models in Cursor?

​Qoder

​Why can't I find the Model option in Qoder?

​OpenClaw

​Why is OpenClaw calling unexpected models?

​Why does OpenClaw connect but hang?

​How to fix OAuth token refresh failures?

​Why does OpenClaw say Unknown model?

​How to handle API rate limits?

​Why does model discovery fail in OpenClaw?

​How do I fix auth.profiles errors in OpenClaw?

​How do I view models configured for Coding Plan in OpenClaw?

​Why do I get "HTTP 401: Incorrect API key provided." in OpenClaw?

​How do I add Coding Plan models without losing OpenClaw config?

​Billing and quota

​When do my quotas reset?

​Does Coding Plan auto-switch to PAYG?

​When does my subscription expire?

​How can I view my usage?

​Can I buy an annual plan?

​Is there a free trial for Coding Plan?

​Can I renew early?

​Can I use coupons when I renew?

​Will I get reminders before my subscription expires?

​Product features

​Which models are supported by Coding Plan?

​What are the model context limits?

​How do I reset or manage my API key?

​Can I manage multiple API keys/IP lists?

​Can I have multiple active subscriptions?

​Can multiple people share one key?

​Which tools are compatible with Coding Plan?

​Are the models full-featured or quantized?

​What should I do if I see data_inspection_failed?

Important resources

Start here: 60-second diagnosis

Most common questions

Why am I still billed after subscribing?

Which API key and base URL should I use?

I got 401 or 403. Is my API key invalid?

Why do I get 404 or connection errors?

Why does the tool say quota exceeded?

Can I use Coding Plan via cURL, Postman, Dify?

Common errors & fixes

How to handle input length limits?

Why is my model "not supported"?

How do I enable thinking mode?

Why is my ModelScope token rejected?

Why is my plan restricted to Coding Agents?

Claude Code

Why am I prompted to migrate from npm?

Why is Claude Code connection failing?

OpenCode

How to resolve "Request Entity Too Large"?

How to fix "Thinking budget" errors?

How do I fix max_tokens exceeded in OpenCode?

Cursor

Why can't I use Coding Plan models in Cursor?

Qoder

Why can't I find the Model option in Qoder?

OpenClaw

Why is OpenClaw calling unexpected models?

Why does OpenClaw connect but hang?

How to fix OAuth token refresh failures?

Why does OpenClaw say Unknown model?

How to handle API rate limits?

Why does model discovery fail in OpenClaw?

How do I fix auth.profiles errors in OpenClaw?

How do I view models configured for Coding Plan in OpenClaw?

Why do I get "HTTP 401: Incorrect API key provided." in OpenClaw?

How do I add Coding Plan models without losing OpenClaw config?

Billing and quota

When do my quotas reset?

Does Coding Plan auto-switch to PAYG?

When does my subscription expire?

How can I view my usage?

Can I buy an annual plan?

Is there a free trial for Coding Plan?

Can I renew early?

Can I use coupons when I renew?

Will I get reminders before my subscription expires?

Product features

Which models are supported by Coding Plan?

What are the model context limits?

How do I reset or manage my API key?

Can I manage multiple API keys/IP lists?

Can I have multiple active subscriptions?

Can multiple people share one key?

Which tools are compatible with Coding Plan?

Are the models full-featured or quantized?

What should I do if I see data_inspection_failed?