Qianwen Cloud: Alibaba Cloud's Agent-Focused MaaS Model Service Platform

Qianwen Cloud is an MaaS model service platform launched by Alibaba Cloud for AI Agents. It aggregates APIs for more than 150 mainstream large models and packages model services into Skills and CLI tools, allowing Agents and developers to complete model selection, invocation, and usage management with a single instruction.

Product positioning

Qianwen Cloud is aimed at AI Agents and developers, consolidating multi-model access and management complexity into a unified entry point. The platform shifts from "serving people" to "serving Agents" and supports text, image, video, and voice modality calls.

Core features

Model Hub and multi-dimensional selection

The platform provides one-stop access to more than 150 mainstream and third-party models, including Qwen, GLM, Kimi, DeepSeek, Wan, and HappyHorse. Users can compare models horizontally by parameter scale, capability dimension, price, context length, modality support, and suitable scenarios, then use real prompts to verify output in the experience page.

Skills packages

The platform packages model selection, authentication configuration, API invocation, and usage query into skill combinations that Agents can understand. An Agent can dynamically route to different models and automatically call vision, image generation, video, and other capabilities without writing code.

CLI command-line tool

The CLI provides login authentication, model query, model invocation, usage query, and environment diagnostics. It supports scripting to automate all model service workflows.

Intelligent usage management

An Agent can pull model usage data in real time, automatically analyze consumption trends, identify anomalies, and output cost optimization suggestions. Process data is transparently presented through CLI logs.

Token Plan subscription

For high-frequency AI coding and Agent tool scenarios, the platform offers subscription packages. A single API Key can be used across all modalities and mainstream AI tools, with controllable monthly fees and significantly lower cost than pay-as-you-go billing.

Recent model updates

The platform has launched several new models: the Qwen3.8 native vision-language series Max model has 2.4 trillion parameters and uses a MoE architecture, described as Tongyi Qianwen's most capable flagship model to date; Qwen3.8-Flash uses a next-generation (Next) model architecture with a total parameter count at the hundred-billion scale and only 6 billion (6B) activated, supports a million-level context window, and is compatible with mainstream OpenAI and Anthropic APIs, so it can connect to tools such as Claude Code and Codex; the official Kimi-K3 API is now live with 2.8 trillion parameters and 1 million token long context, targeting programming and reasoning scenarios.

How to use

1. Visit the official website: open https://www.qianwenai.com/ to browse the Model Hub and compare the models you need.
2. Install CLI: run npm install -g @qianwenai/qianwen-cli in the terminal.
3. Configure Skills: run npx skillsadd QianWen-AI/qianwen-ai to install the Qianwen Cloud skills package for an Agent, or follow the Skills.md documentation provided by the platform.
4. Authenticate and call: complete login authentication through the CLI, obtain an API Key, and use a single instruction to let an Agent automatically complete model selection, invocation, and task execution.
5. Monitor usage: use the CLI to view model usage, billing, and free quota in real time. The Agent will automatically analyze trends and provide optimization suggestions.

Use cases

• AI coding and development: developers can use programming agents such as Qoder to directly call the Qianwen Cloud CLI, automatically completing model selection, code generation, and deployment.
• Multimodal content creation: leveraging Wan capabilities, Agents can automatically schedule text-to-image, text-to-video, and image-to-video models for cinematic-grade content batch production.
• Enterprise smart office: integrate Wukong (悟空) and other work assistants to use one instruction to complete tasks such as schedules, documents, approvals, and meetings across multiple applications.
• Intelligent cockpit and mobility: relying on Qianwen large models, integrate trip assistant, intelligent planning, long-term memory, and other capabilities.
• Social and virtual companionship: support virtual IP creation and real-time emotional dialogue for social human-like interaction scenarios.

Advantages

• Positioned as China's first model service platform with deep Agent orientation, reshaping the core interaction paradigm of model platforms.
• Skills and CLI toolkits are open source on GitHub, so developers can freely extend and integrate them.
• One API Key works with more than 150 models, reducing multi-platform switching and multi-key management.
• Token Plan subscriptions can be as low as 40% off, and one million input tokens can cost as little as CNY 1.
• Supports autonomous Agent invocation while retaining manual online trial, comparative selection, and prompt debugging entry points.

Limitations

• API Keys from different billing methods are not interchangeable: Token Plan Team Edition, Coding Plan, and pay-as-you-go billing have separate API Keys and Base URLs. Using the wrong key will not deduct from the plan quota.
• Token Plan is limited to interactive use in compatible AI coding and agent tools. It cannot be used for automation scripts or application backends. Violations may result in subscription suspension or API Key bans.
• Prices and entitlements change frequently; review the latest rules on the official website before use.

Pricing verification note

Public platform information shows: subscriptions can be as low as 40% off; one million input tokens can cost as little as CNY 1; Qwen3.8-Max is available on Token Plan with individual and enterprise editions from CNY 39; Qoder / Meoo customers can get nightly limited-time 60% off for Qwen 3.8-Max/Flash and 80% off / 50% off for Qwen 3.7-Max/Plus; more than 400 models provide free quota. These prices and promotions may change at any time, so refer to the official page at https://www.qianwenai.com/.

Official resources

• Website: https://www.qianwenai.com/
• CLI installation: npm install -g @qianwenai/qianwen-cli
• Skills configuration: npx skillsadd QianWen-AI/qianwen-ai

FAQ

What is Qianwen Cloud?

Qianwen Cloud is an MaaS model service platform launched by Alibaba Cloud for AI Agents. It aggregates APIs for more than 150 mainstream large models and supports Skills and CLI tools to help Agents and developers complete model selection, invocation, and usage management.

Can the same API Key be used across all compatible tools?

Yes. The same API Key can be used in all compatible AI tools, with shared quota consumption.

Can API Keys from different billing methods or subscriptions be mixed?

No. Token Plan Team Edition, Coding Plan, and pay-as-you-go billing have separate API Keys and Base URLs and must not be mixed. Using another API Key by mistake will not deduct from the Token Plan Team Edition package quota.

Can Token Plan be used for automation scripts or application backends?

No. Token Plan is limited to interactive use in compatible AI coding and agent tools and cannot be used for automation scripts or application backends. Violations may result in subscription suspension or API Key bans.

Do new users get a free quota?

Yes. More than 400 models provide a free quota covering text, image, audio/video, and other multimodal scenarios, available upon registration. Specific quota amounts and scope are subject to the official website.

Alternative solutions

If you need similar multi-model aggregation and Agent toolchain services, you may look for other MaaS platforms that support multi-model API aggregation, Skills/CLI extensions, and subscription billing. Because the source material does not list specific competitor names or comparison conclusions, this article does not add unverified alternatives. It is recommended to evaluate horizontally based on model coverage, Agent compatibility, billing model, and official documentation.