Chat with Qwen3.8-Max

AI ChatQwen3.8-Max

What is Qwen3.8-Max?

Qwen3.8-Max is Alibaba's multimodal flagship for reasoning, coding and extended work. Alibaba's announcement dated August 3, 2026 describes a mixture-of-experts model with 2.4 trillion parameters, 95 billion active, and API access at launch. As checked September 7, QwenCloud lists the September 2 snapshot, qwen3.8-max-0902, also named qwen3.8-max-2026-09-02. The provider describes coding and collaboration improvements. Both it and qwen3.8-max appear in current API documentation; check which version your service actually routes to.

Qwen3.8-Max Capabilities

  • Read visual evidence alongside text: The hosted model accepts text, images and video and returns text, with a 1M-token context window. Use source labels and ask for page or timestamp references. Context capacity does not guarantee complete recall, and Studio upload limits may differ.
  • Choose reasoning and connect tools: QwenCloud's API guide lists thinking, function calling, structured output and built-in tools for both Max IDs. The application must enable tools. Inspect returned evidence and validate JSON before using it downstream.
  • Budget for the full answer: Singapore International API pricing, checked September 7, 2026, is USD $2 per million input tokens and $6 per million output tokens for both IDs. Output billing includes reasoning. Regions, caching and tool charges can change the bill; these are provider rates, not Ottermind prices.

Tasks to Try with Qwen3.8-Max

Give the model a concrete deliverable and a way to check it. Keep the original requirements visible through every revision.

  • Repair a UI from a screenshot

    Supply the screenshot, relevant code and the behavior that fails. Ask for a focused patch and a test, then compare the rendered result with the original. Take the reviewed brief into the AI Website Builder for the next iteration.

  • Reconcile a report with its charts

    Provide the report and chart images. Ask for claims that disagree with the plotted values, including units, dates and source locations. Recalculate key figures and mark unreadable labels as unknown before drafting a correction.

  • Compare supplier proposals

    Identify differences in scope, deadlines and exclusions across the proposals. Require a supporting passage for each finding. Turn verified gaps into questions and refine the supplier reply in the AI Email Generator.

  • Turn a video into an editing brief

    Ask for a timeline of relevant moments with timestamps, a proposed sequence and missing shots. Verify the timestamps against the footage. The model can describe an edit; producing the final video requires a separate editing workflow.

Qwen3.8-Max vs Qwen3.7-Max

Decision pointQwen3.8-MaxQwen3.7-Max
Public QwenCloud inputsText, images and videoText-only interface
Context window1M tokens1M tokens
Singapore API input / outputUSD $2 / $6 per million tokensUSD $2.50 / $7.50 per million tokens
A useful selection testTry a task that needs visual evidence and code togetherKeep as a baseline for a text workflow already tested

Where Qwen3.8-Max Fits

Reasons to evaluate it

  • A visual upgrade over 3.7: Qwen3.7-Max's product page describes a text-only public interface. For screenshot debugging or chart review, 3.8 can take the visual source directly. Compare on the same deliverable instead of assuming every text task will improve.
  • Plan, act and check in stages: Alibaba reports a 16-day internal engineering demonstration in its launch announcement. Treat that as a vendor experiment, not a completion guarantee. Try one bounded milestone with tool logs and acceptance tests before extending the run.

Checks that still matter

  • Watch for quiet scope changes: A polished result can omit a difficult requirement. Require a final list of completed, failed and untested items tied to your original brief. Test behavior yourself rather than accepting a statement that all checks passed.
  • Measure cost per accepted result: Lower token prices can be offset by long reasoning, repeated calls and corrections. Compare the same task with a time limit and token budget. Record correctness, elapsed time and retries before choosing a default.

What Users Report

The discussions below mix release and older preview experiences. They are individual reports, not a consensus, controlled benchmark or evidence that the 0902 update fixes every issue.

Speed and token use divide opinions

A Qwen community discussion praises coding speed and token economy, while replies challenge both. Some comments concern Preview. Reproduce your workload with the exact version and provider before relying on a claimed saving.

Preview sometimes simplified the assignment

One preview reviewer reports that a poster task replaced required photos with typography. Replies include different experiences. Keep this as an instruction-following test for the current snapshot, not a measured failure rate.

A published ID may not match a plan

A September access report describes a model-not-found error for 0902 on a token plan. A reply suggests automatic routing, but that is not official confirmation. Check the plan, endpoint and supported IDs.

Start a Qwen3.8-Max Task in Ottermind

1

Write the result you need

Enter the objective, source material and acceptance criteria above. Include constraints that must survive every revision, such as required images, dates or supported behavior.

2

Check the model in Studio

Continue to Studio and sign in if needed. Select Qwen3.8-Max if available to your account, then check files, tools and usage terms. This page does not automatically select or provision the model.

3

Review before continuing

Inspect the first result against its sources and run relevant checks. Request a specific correction, keep the accepted version and carry it into the next connected task.

Qwen3.8-Max FAQ

Is Qwen3.8-Max the same as 0902?

0902 is a dated upgrade within Qwen3.8-Max. QwenCloud names it qwen3.8-max-0902, with alias qwen3.8-max-2026-09-02. Both the base and dated IDs are documented as of September 7, 2026; verify routing with your provider rather than assuming they always resolve identically.

Can I download Qwen3.8-Max weights?

Qwen's official weight repository is named Qwen3.8-2.4T-A95B. Its card distinguishes the managed Qwen3.8-Max service, which adds vision, non-thinking support and built-in tools. Do not assume the hosted ID is an identical downloadable model or carries another Qwen model's license.

Does it create images or finished videos?

The hosted model's documented output is text. Visual input lets it analyze footage or describe an image; media generation and editing need separate tools. Check what the chosen Studio workflow supports.

Are these rates my Ottermind bill?

No. The quoted rates are Alibaba Cloud's Singapore International API prices. Ottermind availability, usage limits and charges depend on your account. Continuing to Studio transfers your prompt; it does not establish model access.

Bring the evidence, define the result

Develop a brief, inspect the answer and refine the next version in Ottermind Studio.