Alibaba Cloud API (Qwen models): pricing, free tier, rate limits and how to start

Use Alibaba Cloud Model Studio to connect a Qwen or other hosted model to an application. Keep workspace access, endpoint selection and the model’s actual billing scope together.

What Alibaba offers through the API

Model Studio offers Qwen and third-party models across text, image, audio and video, with dedicated entries for embeddings and reranking. Official documentation.

Choose the model task before choosing a client example. A text completion, an image generation and a transcription are different application operations, even when the same cloud account owns them. Start with the endpoint appropriate to the required input and output, then inspect its request contract and charge categories.

The Qwen quickstart documents both DashScope and OpenAI-compatible client paths, with the API host tied to the selected workspace access configuration. Official documentation.

The workspace is therefore part of the integration record. Keep the key, compatible-mode base URL and exact model identifier together in protected configuration. A generic endpoint copied from an older example can fail even when the Python call looks correct. Use the host shown by the current console and documentation for the intended service scope.

The Qwen catalog combines general-purpose candidates with models aimed at coding, visual understanding and other specialized work. Treat the family description as a way to construct an evaluation shortlist. It does not establish that a particular candidate will meet the output checks of your application. A code model should be evaluated on accepted patches; an extraction model should be evaluated on accurate fields tied to source evidence.

For a first integration, use a safe fixture that can be inspected completely. Keep a short document, a clear extraction request and a rule for missing information. If the task involves media, choose an example whose relevant content is known and verify the returned result manually before creating automated evaluation checks. Record the request interface and model identity with the outcome.

Account organization becomes important when the prototype is shared. Decide which workspace owns the application, who manages its key and which billing view will be used to reconcile usage. Keep the stopping condition explicit before adding background jobs. A cloud account can contain many unrelated services, so an unexplained balance movement is not a useful application metric.

This hub connects the model and operational questions. The Models page separates catalog roles and version choices. Pricing explains billable categories and context conditions. The API-key page covers the current host-and-key workflow. Free tier explains model-specific offer scope, while Limits and Errors distinguish account pacing from billing and request failures.

The Python tutorial supplies a small native DashScope request without inventing a workspace identifier. After it works in your account, extend it one requirement at a time. Keep a passing baseline, measure the returned usage and inspect completion state. That sequence makes a workspace migration or model replacement easier to evaluate later.

Before deploying, save a concise operating note with the model, interface, workspace owner, secret location and accepted fixtures. Include where to inspect usage and how to stop scheduled calls. The note should contain enough context to maintain the service while excluding the credential itself.

Pricing at a glance

Read Model Studio pricing for model categories and applicable conditions.

Models available

Inspect Qwen and hosted-model families.

Free tier and how to get a key

Check model-specific free quota and workspace key setup.

Rate limits and tiers

Plan shared account traffic using Model Studio limits.

Service status

Start in 5 minutes

Follow the Alibaba Qwen Python tutorial after copying the correct native DashScope URL.

import os
import dashscope
from dashscope import MultiModalConversation
dashscope.base_http_api_url = os.environ["DASHSCOPE_HTTP_BASE_URL"]
result = MultiModalConversation.call(api_key=os.environ["DASHSCOPE_API_KEY"], model="qwen3.8-max", messages=[{"role":"user","content":[{"text":"Hello"}]}], request_timeout=60)

Recent changes

Use the AI API cost calculator to turn the model and workload you are considering into an estimate.

Last verified 12 Sep 2026 · Source ↗

Frequently asked questions

Sources

• First Qwen API call ↗
• Recommended models ↗
• Model inference pricing ↗
• API keys and permissions ↗
• Rate limits ↗
• Error codes ↗
• New-user free quota ↗
• Model usage ↗
• Context cache ↗
• Batch API ↗
• Billing and cost management ↗
• Dynamic rate limiting ↗
• Qwen Coder ↗
• Model updates ↗

Last verified 12 Sep 2026 · Source ↗