Direct answer
Qwen 2.5 Max is a text-in, text-out Qwen model whose original API ID is qwen-max-2025-01-25. Alibaba Cloud Model Studio currently documents a rolling-updated qwen-max entry and says that version is functionally equivalent to the dated snapshot. Because the alias is rolling, it is not a fixed-version choice; the documentation does not establish that the dated ID remains available through every region and workspace route. (Model Studio model documentation)
For the currently documented qwen-max entry, Model Studio lists a 32,768-token context window, a 30,720-token maximum input, and an 8,192-token maximum output. The page also describes the entry as functionally equivalent to the dated snapshot; treat these limits as the provider’s listing for qwen-max, not as a separately confirmed, immutable specification for every historical snapshot route. (Model Studio model documentation)
What it can—and cannot—tell you about fit
Model Studio describes qwen-max as a large Qwen2.5-series language model and lists support for Chinese, English, and other languages. That is useful for narrowing a shortlist, but it does not establish how the model will perform on your prompts, data, latency target, or budget. Test with representative workloads rather than treating vendor descriptions as a guarantee. (Model Studio model documentation)
| Requirement | What Model Studio documents for qwen-max |
Selection implication |
|---|---|---|
| Text chat or generation | Text input and text output | Suitable to evaluate for text-only workflows; test with your own prompts. |
| Image or video input | The listed input modality is text | If image or video input is required, select a model whose documentation lists that modality. (Model Studio model documentation) |
| Function calling or web search | Listed as supported for China (Beijing) and Singapore; unsupported in the International scope | Check the serving scope before choosing if either feature is required. (Model Studio model documentation) |
| Structured output | Listed as supported for China (Beijing), Singapore, and International scope | The capability listing does not specify output-mode restrictions; verify that your required schema and client mode are supported. (Model Studio model documentation) |
| Prefix completion, context caching, or batch inference | Each is listed as supported for China (Beijing), Singapore, and International scope. Fine-tuning is listed as unsupported in those scopes. | Confirm that the documented feature and your intended route meet your use case. (Model Studio model documentation) |
The provider’s capability and pricing tables distinguish China (Beijing)/Singapore from an International scope. Do not assume a feature, price, or rate limit listed for one scope applies to another. The documentation does not establish task-specific performance for your workload, so compare candidate models using the same prompts and evaluation criteria. (Model Studio model documentation)
Production API: a documented route, with a model-ID distinction
Qwen’s launch post identifies qwen-max-2025-01-25 as the original API model name. Model Studio’s current model page documents the rolling qwen-max entry and says it is functionally equivalent to that snapshot. The Singapore API documentation supplies a workspace-specific OpenAI-compatible base URL pattern, but the available documentation here does not establish that the dated snapshot ID is currently enabled through that exact Singapore route and billing plan. The example below therefore uses the documented rolling alias, not the fixed snapshot. (Model Studio model documentation)
For a Singapore workspace, Model Studio documents this base URL pattern: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1. The chat-completions path is /chat/completions. Use a key and base URL that match your region and billing plan, and replace {WorkspaceId} with your workspace ID. (Model Studio compatibility documentation)
A minimal Python example using the documented rolling alias:
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["DASHSCOPE_API_KEY"],
base_url="https://YOUR_WORKSPACE_ID.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
)
response = client.chat.completions.create(
model="qwen-max",
messages=[
{"role": "user", "content": "Write a one-sentence summary of this paragraph."}
],
)
print(response.choices[0].message.content)
Install the openai Python package and set DASHSCOPE_API_KEY to your Model Studio API key. Do not place a secret key in source code. This is an illustration of the documented compatible-interface pattern, not a claim that the request was executed here. If you specifically need the dated qwen-max-2025-01-25 snapshot, confirm that exact ID is enabled for your workspace and route before substituting it in the example. (Model Studio first-call guide; compatibility documentation)
Pricing and operational trade-offs
Model Studio’s currently listed original API-call prices for the rolling qwen-max entry are CNY 2.4 per million input tokens and CNY 9.6 per million output tokens for China (Beijing)/Singapore. Its separately labeled International scope lists CNY 11.743 per million input tokens and CNY 46.971 per million output tokens. The page says these are original prices and excludes limited-time promotions. These figures apply to the documented qwen-max listing; they do not by themselves confirm pricing for a historical snapshot on a particular account or route. Check the applicable scope and account terms before budgeting. (Model Studio model documentation)
For illustration, a request with 10,000 input tokens and 2,000 output tokens comes to CNY 0.0432 at the listed China (Beijing)/Singapore rates: (10,000 × 2.4 + 2,000 × 9.6) ÷ 1,000,000. At the listed International rates, the same token counts come to CNY 0.211372. These are arithmetic estimates from the published per-token rates, not a quote for a specific account. (Model Studio model documentation)
The same page lists rate limits of 1,200 requests per minute and 1,000,000 tokens per minute for China (Beijing)/Singapore, and 600 requests per minute and 1,000,000 tokens per minute for the International scope. Confirm which limits apply to your account and route before relying on them for capacity planning. (Model Studio model documentation)
Bottom line
Consider Qwen 2.5 Max when a text-only model fits your task and you can evaluate it against alternatives using your own workload. Model Studio’s current qwen-max listing provides useful limits, region-specific feature support, prices, and rate limits, but it is rolling-updated. The original snapshot ID is qwen-max-2025-01-25; current availability of that exact ID through the Singapore example route is not established here. Before production adoption, verify the model ID enabled for your workspace, region and billing plan, and the limits and price that apply to that route. (Model Studio model documentation)
Sources
Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model
qwen-max Model Info — Alibaba Cloud Model Studio
Call Qwen models via OpenAI API — Alibaba Cloud Model Studio
Make your first API call to Qwen — Alibaba Cloud Model Studio