If you mean the Qwen Coder website, go to coder.qwen.ai. Qwen describes it as a coding agent that runs in a cloud sandbox and can write features, fix bugs, and export work to GitHub. Its public description does not identify the exact model powering each task or document an API for the portal. For a coding agent in a local project, use Qwen Code, the separate open-source terminal and IDE agent. For an application integration, Alibaba Cloud Model Studio provides a distinct model API route. (Qwen Coder; Qwen Code; Model Studio model catalog)
Which Qwen coding route should you choose?
| What you want to do | Use this route | What it means |
|---|---|---|
| Try an online coding agent | Qwen Coder | A browser-based coding agent with a cloud sandbox and GitHub export. Its public description does not specify a model ID or API contract. |
| Work with files in your development environment | Qwen Code | An open-source terminal-first agent with editor integrations. It is software for using models, not a model name. |
| Integrate a Qwen model into an application | Alibaba Cloud Model Studio | A hosted model catalog with API documentation. Check the selected model’s deployment details and billing. |
Is Qwen Coder the same thing as Qwen3-Coder?
Not necessarily. Qwen Coder is the name of the web coding-agent destination; Qwen Code is the open-source coding agent for a terminal or supported IDE; Qwen3-Coder is a family of coding models. Qwen’s model repository lists variants including Qwen3-Coder-480B-A35B-Instruct, Qwen3-Coder-30B-A3B-Instruct, and Qwen3-Coder-Next. The Qwen Coder portal page does not say which model, if any, powers a particular portal session. Don’t treat the portal name as a model ID. (Qwen Coder; Qwen3-Coder repository)
A model card or checkpoint describes a model; it does not by itself establish that a hosted portal or API serves that exact model, version, context window, or feature set.
A documented hosted coding model: Qwen3.8-Max on Model Studio
Alibaba Cloud Model Studio identifies Qwen3.8-Max with API model ID qwen3.8-max and lists snapshot qwen3.8-max-0902 (alias qwen3.8-max-2026-09-02). Its model documentation describes improved coding capability and autonomous planning and iteration for long-horizon tasks. That makes it a candidate to evaluate for substantial, multi-step engineering work—not proof that it will complete a project autonomously or outperform another model. The documented API route also does not establish which model powers the Qwen Coder website. (Model Studio Qwen3.8-Max specifications)
The model information lists a 1,000,000-token context window, with maximum input and output lengths of 991,808 and 131,072 tokens respectively. These are model-page limits, not a promise that every API request can use the full input window. For the Responses API, Model Studio says usable input context is approximately 80% of the model window to reserve room for built-in tools and reasoning; the actual usable length can vary with the request configuration. (Qwen3.8-Max model information; Responses API documentation)
The documentation describes intended capabilities, not measured performance on your codebase. Before adopting it, test representative work—such as fixing a failing test, changing several related files, or following project conventions—and review its proposed changes. Treat coding quality and tool-assisted completion as things to validate for your own workflow.
Production API: a documented Singapore route
Model Studio lists Qwen3.8-Max for Singapore and includes qwen3.8-max among the models supported by its OpenAI-compatible Responses API. For that regional route, the documented workspace-specific base URL is https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com; the Responses endpoint appends /compatible-mode/v1/responses. Replace {WorkspaceId} with the ID from your Model Studio workspace. The model ID is qwen3.8-max. (Model information; Responses API compatibility and regional domains)
You need a Singapore-region Model Studio API key and the matching workspace. Model Studio says regional keys and endpoints cannot be used across regions; its API-key setup lets you select a workspace and permissions. Send the key as a bearer credential and provide the model ID and an input value in the request. Keep the key confidential. Confirm the current API documentation and console settings before deploying, especially if your account uses a Token Plan or Coding Plan key rather than a pay-as-you-go key. (API-key setup; Responses API documentation)
API pricing and purchase fit
The Model Studio standard-price table lists Singapore pricing for qwen3.8-max at $2 per million input tokens and $6 per million output tokens, for requests with up to 1 million input tokens. These are standard pay-as-you-go rates, not an account-specific quote; promotions may differ. Cached input can have separate pricing, so check the console and the current cache-billing rules if you plan to use caching. (Model Studio model inference pricing)
Those rates apply to the documented Singapore route, not automatically to every region, portal, or subscription. The Qwen Coder portal’s public description does not establish its pricing. Compare your expected input and generated output usage against the API rates, and account for the Responses API’s smaller practical input allowance when tool calls or reasoning are involved.
Bottom line
- Want the online coding portal? Start at Qwen Coder. Its public description does not identify the model behind a session.
- Want an agent working from your terminal or IDE? See Qwen Code.
- Want an API for an application? Model Studio documents Qwen3.8-Max as
qwen3.8-maxthrough its Responses API, including a Singapore workspace-specific route. Use a region-matched key and workspace, and account for the API’s usable-context qualification. - Need a coding model candidate? Qwen3.8-Max is worth evaluating for multi-step engineering work based on the vendor’s coding and long-horizon planning descriptions; those descriptions are not independent performance results.
- Need the exact model behind the Qwen Coder portal? Its public description leaves that unspecified; do not assume it is Qwen3-Coder or Qwen3.8-Max.
For a broader overview of API access and deployment choices, see Qwen API: Access, Evaluation, and Deployment Options.
Sources
Alibaba Cloud Model Studio model catalog
Qwen3.8-Max model information and pricing