Qianwen AI Platform

What is the Qianwen AI platform?

The Qianwen AI platform is a full-modal large model service and application development platform designed for developers and enterprises.

It offers model APIs, key management, usage billing, agent tools, and the capability to develop AI-native applications.

Differences from Qianwen AI assistant

The three names are related to each other, but their methods of use, billing models, service responsibilities, and open-source scope differ.

Core competencies

• Text generation, reasoning, and long-context understanding.
• Understanding of image and video content.
• Generation of images, videos, and audio.
• Speech recognition and real-time voice dialogue.
• Vectorization, reordering, and knowledge retrieval.
• Function calls, MCP, and agent tools.
• API Key, billing, and usage analysis.

Full-modal model market

The platform model market includes the Qwen series as well as other models that are available; it is possible to select different levels of performance and costs depending on the task at hand.

API and protocol compatibility

The Qianwen AI platform offers a unified API and is compatible with common invocation protocols such as those of OpenAI and Anthropic.

• Use an API Key to complete authentication.
• It supports stream output and structured results.
• Supports function calls and agent tools.
• It can be integrated with mainstream programming and Agent clients.
• The models, endpoints, and parameters should be based on the documentation.

Developer Dashboard

The workspace is used to manage keys, business spaces, model calls, usage metrics, and billing records.

• Create and rotate API keys.
• Distinguish between development, testing, and production environments.
• View the Token, number of requests, and latency.
• Monitor success rates and cost changes.
• Isolate different projects by business area.
• Set budgets and abnormality alerts.

Agents and tools

The platform offers capabilities such as online searching, web scraping, code interpreters, image search, function calling, and MCP.

• Calling tools may incur costs related to the model and the tools themselves.
• External results may be outdated or contain errors.
• Tool permissions should be configured to a minimum required for each task.
• Write, payment, and deletion operations require confirmation.
• All critical calls should have audit logs recorded.

Which AI development tools are supported?

The Token Plan can be used in conjunction with a variety of programming and agent tools that are compatible with the OpenAI or Anthropic protocols.

• Qwen Code.
• Claude Code and Codex.
• Cursor, Cline, and OpenCode.
• Kilo CLI and OpenClaw.
• Other clients that support custom endpoints.

The functionality, privacy policies, and additional costs of third-party tools are the responsibility of their respective providers.

Guide to Using the Qianwen AI Platform

1. Log in to the platform using an Alibaba Cloud account.
2. Read and agree to the service agreement and privacy policy.
3. Confirm the free quota and validity period for new users.
4. Create a separate business space.
5. Choose Pay-as-you-go or Token Plan.
6. Create an API Key and save it securely right away.
7. Select the appropriate model and the correct server endpoint.
8. Use the official example to make the first call.
9. Set timeouts, retries, and rate limits.
10. Monitor tokens, tool calls, and costs.
11. Add manual review for high-risk outputs.
12. Regularly rotate keys and check bills.

Price of the Personal Version of Token Plan

The Token Plan is measured in Credits, with a fixed limit applied on a 7-day basis.

The price was verified on August 30, 2026; limited-time discounts, supported models, and quota rules are as specified on the subscription page.

How is the quota for the Token Plan calculated?

Different models and tools consume Credits according to their own conversion rules; Credits cannot be simply equated with the number of Tokens.

• The limit is reset every 7 days.
• The service will be suspended once the window limit is reached.
• The available credit is restored once the new window is opened.
• Different models consume different amounts of Credits.
• Networking and knowledge tools may incur additional costs.
• Check the list of currently supported models before subscribing.

Pay-as-you-go

Pay-as-you-go pricing is based on the actual number of successful API calls; there is no need to purchase fixed packages in advance.

Failed requests do not incur any model costs, nor do they consume the corresponding free quota.

Examples of prices for some models

The unit price of the model changes rapidly; the table below is intended solely to help understand the billing scale shown by the official source.

The prices listed above were verified on August 30, 2026; the final amount will be based on the details of the selected model and the invoice.

Price of built-in tools

The prices of the tools were verified on August 30, 2026; the exemption status and pricing rules may change.

Free quota

The platform offers new users free access to various models, but the specific quantity, validity period, and scope of application vary.

• First, check the available credit amount on the account page.
• Record the expiration date for each quota.
• After the free quota is used up, it may switch to a paid version.
• Create budget alerts to avoid unexpected expenses.
• Production applications should not rely on temporarily granted quotas.

How to choose the billing method

Cost control methods

• Select the appropriate model based on the task’s complexity.
• Limit the maximum number of input and output tokens.
• Reuse stable prompts and context caching.
• Non-real-time tasks use the Batch API.
• Avoid bringing the entire history into the conversation indefinitely.
• Set a maximum limit on the number of tool calls.
• Split keys and budgets by project.

API Key security

The official documentation explicitly advises against hard-coding keys in the source code, as well as against committing keys to version control systems.

• Use environment variables or key management services.
• Different keys are used for development and production.
• Grant access rights only to the services that are necessary.
• Upon detecting a leak, it is immediately revoked and replaced.
• Do not expose server keys in the browser frontend.
• Regularly review the sources of calls and abnormal costs.

Data and content security

When invoking the model, business documents, user inputs, images, audio, and tool outputs may be transmitted.

• Classify and mask the data before uploading.
• Do not send passwords or long-term credentials.
• Confirm the data processing terms for the selected model.
• Manual approval is set up for sensitive transactions.
• Use content moderation and input/output protection.
• Handle personal and cross-border data in accordance with the law.

Stability and production deployment

• Set reasonable timeouts and retry options for the interface.
• Exponential backoff is used for rate limiting.
• Record the model, version, and request identifier.
• Prepare fallback plans for critical processes.
• Strict verification is performed on structured output.
• Do not allow the model to carry out high-risk operations directly.

Open-source status

The Qianwen AI platform hosts MaaS services; it is not an open-source software project.

Projects such as Qwen Model, Qwen Code, and Qwen Agent can be found on the official GitHub page, but the licenses for each of them need to be checked separately.

• The openness of a model does not equate to the openness of a cloud platform.
• Code licensing is not the same as model weight licensing.
• API calls are subject to the platform service agreement.
• Third-party models are governed by their respective terms.

Which users are it suitable for

• Teams responsible for developing chat, search, and knowledge base functions.
• Developers who create programming and automation agents.
• Products that require image, video, and voice APIs.
• Companies that wish to manage multiple models in a unified manner.
• Teams that need to manage their budgets on a pay-as-you-go basis or through subscriptions.
• Applications that need to be compatible with mainstream SDK protocols.

Product advantages

• Covers text, images, videos, and audio.
• Unified APIs and comprehensive development documentation are provided.
• Compatible with common programming and Agent tools.
• It supports both subscription and pay-as-you-go models.
• Provides key, billing, and usage management.
• Supports Batch, caching, and various tools.

Usage restrictions and precautions

• Model prices and promotional discounts change rapidly.
• Credits cannot be equated directly with Tokens.
• The free quota has limits in terms of scope and validity period.
• Third-party models have their own separate terms of service.
• Calling tools may incur additional fees.
• The model’s output may be incorrect or incomplete.
• Production systems must ensure security and be capable of operating in a degraded mode.

Frequently Asked Questions

Is the Qianwen AI platform the same as the Qianwen AI assistant?

They are different: the former provides model APIs for developers, while the latter offers an AI assistant experience primarily for individual users.

Can the Qianwen AI platform be used for free?

New users are entitled to a certain number of free model instances; the exact number, validity period, and applicable models are specified on the benefits page.

Do the Credits in the Token Plan keep accumulating?

The personal version has a fixed quota allocated every 7 days; once this limit is reached, usage is suspended, and the quota is reset when a new period begins.

Which API protocols is the Qianwen AI platform compatible with?

The platform offers official APIs and is compatible with common interface protocols and development tools such as those of OpenAI and Anthropic.

Is there a fee for failed API calls?

The official billing documentation states that failed requests do not incur any costs related to the models, nor do they consume any of the free quota allocated.

How to prevent API keys from being leaked?

Do not hard-code keys in the source code or frontend; instead, use environment variables or key management services and rotate them regularly.

Is the Qianwen AI platform an open-source project?

The platform itself is not open source; the public projects related to the Qwen model and the development tools must be used in accordance with their respective licenses.