SuperQode
Alibaba Cloud Model Studio
Qwen4-Max
macOS

Frontier Intelligence with Alibaba Cloud Model Studio and SuperQode

Use Qwen4-Max as the model behind SuperQode's own harness on a Mac, with a pay-as-you-go key and your region's endpoint.

November 18, 2026
7 min read
By Superagentic AI
Frontier Intelligence with Alibaba Cloud Model Studio and SuperQode

SuperQode • Alibaba Cloud Model Studio • Qwen4-Max

Frontier Intelligence with Alibaba Cloud Model Studio and SuperQode

Alibaba Cloud Model Studio now serves Qwen4-Max through its OpenAI-compatible API. We wanted to know what it takes to use it as the model behind SuperQode's own harness on a Mac: a pay-as-you-go key, the endpoint for the region you chose, and nothing else. SuperQode keeps the tools, the approval prompts and the edits on your Mac. Each request appears in Model Studio's audit log, and the changed file on your Mac is the other half of the record.

We tried it on the Singapore international endpoint with a pay-as-you-go key. The workspace catalog listed 103 models, qwen4-max among them. The recording shows the connection, a one-line edit in a small project, and the same call in the Singapore audit log.

Download SuperQode for Mac · Qwen setup guide

Watch the demo

Watch Qwen4-Max in SuperQode on YouTube

Connect from the Mac app

Open a project in SuperQode. On a new thread, choose API key, then Alibaba Model Studio. The route starts on Your region. Leave it there for a pay-as-you-go key from an international region: Singapore, US (Virginia), Germany (Frankfurt), Japan (Tokyo) or China (Hong Kong). A China key uses the China route below.

In the Model Studio console, pick the region, create a pay-as-you-go API key there, and copy the OpenAI-compatible base URL from the same console. Paste the key, then paste the URL into Model Studio endpoint. Remember in Keychain is on by default, so the key is saved in your Mac's Keychain rather than in a settings file.

For the Singapore workspace we used:

Endpoint
https://dashscope-intl.aliyuncs.com/compatible-mode/v1

A dedicated workspace URL looks like https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1. For EU data residency, Model Studio offers Germany (Frankfurt). It uses workspace URLs only, in the form https://{WorkspaceId}.eu-central-1.maas.aliyuncs.com/compatible-mode/v1. Choose the EU deployment scope when you create the workspace. The Singapore URL above is the one in this recording.

SuperQode does not choose a region for you. The URL you paste is the region, and a key only works with the region that issued it. See base URLs by region for the full list.

The endpoint field accepts only an official Model Studio host, either a *.maas.aliyuncs.com workspace host or a DashScope host, with the path /compatible-mode/v1. Anything else is refused, so the key is not sent to another server. Token Plan and Coding Plan hosts are refused here too, because they belong to their own routes. Use provider stays off until the endpoint is set.

Click Load workspace models. SuperQode asks the workspace for its catalog and lists the model IDs it returns. This does not save the key and does not send a prompt. The list can show built-in examples before the workspace answers. Wait until the caption says Loaded N workspace models, then select qwen4-max and choose Use provider.

Why thinking starts at low

Qwen4-Max thinks at a high effort by default, and that thinking is billed. When you have not set an effort, SuperQode sends low for qwen4-max, so a first trial turn does not spend the large default budget. An effort you set explicitly is passed through, so you can raise it when a task needs more.

Thinking and the answer also share one output budget. If a request asks for a short output cap, SuperQode raises it for Qwen4-Max, so the model does not use the whole allowance on thinking and return an empty reply.

China, Token Plan and Coding Plan

The China route follows the same pay-as-you-go flow with a key from the China console. Leave the endpoint empty to use https://dashscope.aliyuncs.com/compatible-mode/v1, or paste a China workspace URL.

Token Plan is a separate route in the same picker. It takes an sk-sp- key and uses Alibaba's fixed Token Plan endpoint, for interactive coding only. If you paste an sk-sp- key on the pay-as-you-go route, SuperQode switches to Token Plan instead of saving it as pay-as-you-go. Coding Plan is set up through Qwen Code, described below, and does not include qwen4-max.

Prices and free quota stay in the Model Studio console, because Alibaba can change them.

One small edit, and the audit log

We opened a folder with a single hello.py that prints hello, and sent:

Prompt
Add this comment as the first line of hello.py: # Qwen4-Max on Model Studio. Then stop.

SuperQode asked for approval before the edit. Afterwards the file started with that comment, and the model chip in the composer still showed qwen4-max.

In Model Studio, the call is under Operations, then Model Monitoring, then View Logs, on the Audit Log tab. Filter to qwen4-max and the last hour. The row shows the request time, the model, the token count, the duration and the status. It does not include the prompt or the reply. The changed file on your Mac is the other half of the record.

The same connection in the terminal

The Mac app and the terminal use the same qwen-intl provider. Set the key and the endpoint from your console, then list the workspace catalog and connect:

Shell
export QWEN_INTL_API_KEY=your-key-here
export QWEN_INTL_API_BASE=https://dashscope-intl.aliyuncs.com/compatible-mode/v1
superqode models --provider qwen-intl --live

Then, inside SuperQode:

Terminal
:connect byok qwen-intl qwen4-max

--live queries the workspace for its own catalog. If QWEN_INTL_API_BASE is unset, SuperQode uses the shared Singapore host above. Set the variable when your console shows a different region. For China, use QWEN_CN_API_KEY and qwen-cn. The regional API keys section of the guide covers both.

Qwen Code, when you want Qwen's own agent

SuperQode can also run Qwen Code over the Agent Client Protocol. That connection keeps Qwen Code's own agent loop, model selection and billing. In the Mac app, open Qwen Code's connection, install the agent if needed, and choose Configure Qwen Code…. In the Qwen Code terminal, run /auth and choose Alibaba Model Studio, your region, and your plan or API key. Exit Qwen Code, let SuperQode check the connection, then choose Use this connection. In the terminal:

Terminal
:connect qwen-code

The two paths answer different needs. Use Qwen Code when you want that agent to own the tools and the plan. Use Alibaba Model Studio with qwen4-max when you want SuperQode's harness to own the tools, the approvals and the project on your Mac. Qwen Code setup in SuperQode

What's next

A workspace host such as https://{WorkspaceId}.{region}.maas.aliyuncs.com returns context limits and tool support with each model. The shared Singapore host in this recording returns model IDs. The low thinking default for qwen4-max is SuperQode's own setting. Next, we want the Mac model list to show the context and tool fields when the workspace catalog returns them, and the connected provider to link straight to that workspace's audit log.

Try it

Open a project in SuperQode, choose Alibaba Model Studio, paste a pay-as-you-go key and the endpoint for your region, and select qwen4-max after Load workspace models reports the live list. Ask for one small edit, approve it, and find the matching row in the audit log.

Model Studio overview · Get an API key · Base URLs by region · Qwen4-Max

The app supports macOS 14 or later on Apple Silicon. Download for Mac · Python package on PyPI · Source on GitHub

📚 Our blogs are also published on

Follow along wherever you already read

💡 Found this helpful? Share it with your network and help others discover these insights!

Try it

Run Qwen4-Max in SuperQode

Download SuperQode for Mac, connect Alibaba Model Studio, and make one small edit with qwen4-max.