Quick answer
Alibaba released Qwen3.8-Max-0902 on September 2 as an upgraded snapshot of Qwen3.8-Max. Qwen says the update improves coding, long-running autonomous development, multi-tool agent work, charts, document parsing, and multimodal perception. The hosted model accepts text, images, and video, returns text, retains a 1-million-token context window and thinking mode, and supports function calling and structured output. QwenCloud lists the dated model ID for API use, but the exact model is not verified in Chat AI's current model directory.
Qwen3.8-Max-0902 is a dated upgrade, not a new open-weight family
Alibaba documents Qwen3.8-Max-0902 as an upgraded snapshot of Qwen3.8-Max and also lists the alias qwen3.8-max-2026-09-02. The official pages present it as a hosted model for QwenCloud and Alibaba Cloud Model Studio; they do not announce downloadable weights for this snapshot. That distinction matters for teams comparing it with Qwen's open-weight releases, because access and deployment choices are different.
Sources: QwenCloud, Alibaba Cloud Model Studio
The release targets coding and longer agent workflows
Qwen says the snapshot is designed for more complex engineering projects, long-horizon autonomous development, multi-tool orchestration, and end-to-end task delivery. Those are provider descriptions rather than independently reproduced performance results. Teams evaluating the update should compare it with the base qwen3.8-max on their own repositories, tool definitions, permission boundaries, and stopping conditions.
Sources: QwenCloud, Alibaba Cloud Model Studio
Vision changes cover charts, documents, and video input
The official model page says native vision understanding has been refined for chart reasoning, document parsing, and broader multimodal perception. QwenCloud lists text, image, and video as supported inputs, with text as the output. This makes the release relevant to document and media workflows, but the announcement does not provide a reproducible vision benchmark for the 0902 snapshot, so users should test the specific file formats and visual tasks they rely on.
Sources: QwenCloud, Alibaba Cloud Model Studio
The 1-million-token context and tool features remain available
Qwen lists a 1-million-token context window and thinking mode for Qwen3.8-Max-0902. The model directory also marks function calling, built-in tools, and structured output as supported. A large advertised context window does not by itself establish reliable retrieval across every position or workload, so long-document and large-codebase evaluations should measure citation accuracy, missed constraints, latency, and total token use.
API access uses the dated model ID
QwenCloud's API example uses qwen3.8-max-0902, while Alibaba Cloud's family page lists qwen3.8-max as the rolling model ID and places 0902 under snapshot versions. Use the exact identifier documented for the endpoint and region you call rather than assuming the rolling alias and dated snapshot are interchangeable. Regional capabilities and rate limits can differ, so confirm the current service documentation before production rollout.
Sources: QwenCloud, Alibaba Cloud Model Studio
QwenCloud lists $2 input and $6 output per million tokens
At publication time, the QwenCloud model page lists $2 per million input tokens and $6 per million output tokens for Qwen3.8-Max-0902, with separate cache prices. Pricing can change and may vary by service or region, so treat those figures as a timestamped QwenCloud listing rather than a universal quote. For agent workflows, measure the full run cost because retries, tool results, reasoning, and long context can matter more than the headline token rate.
Sources: QwenCloud
Run a controlled evaluation before replacing the base model
Compare the snapshot and the rolling qwen3.8-max alias on the same task set, prompts, tools, context, temperature, and stopping rules. For coding, track test pass rate, regressions, tool calls, elapsed time, and review effort. For documents and vision, track extraction accuracy, grounding, and failures on tables or long files. Also record the exact model ID returned by the provider so results remain attributable after a rolling alias changes.
Sources: QwenCloud, Alibaba Cloud Model Studio
Qwen3.8-Max-0902 availability in Chat AI is unverified
The exact Qwen3.8-Max-0902 model does not appear in Chat AI's current public model directory, so this article does not claim that it is available in Chat AI. Consult the model directory and in-app selector for currently verified models; a future listing would be a separate availability update.
Sources: Chat AI
Frequently asked questions
What readers usually ask
What is Qwen3.8-Max-0902?
It is Alibaba's September 2 dated snapshot of Qwen3.8-Max, with provider-reported improvements for coding, long-running agent work, multi-tool orchestration, and vision tasks.
Is Qwen3.8-Max-0902 an open-weight model?
The official release pages describe hosted access through QwenCloud and Alibaba Cloud Model Studio and do not announce downloadable weights for this snapshot.
What context window does Qwen3.8-Max-0902 support?
QwenCloud lists a 1-million-token context window. Actual long-context quality, latency, and cost still depend on the workload and service configuration.
What is the Qwen3.8-Max-0902 API model ID?
QwenCloud documents qwen3.8-max-0902 and the alias qwen3.8-max-2026-09-02. Confirm the identifier supported by your endpoint and region before deploying it.
How much does Qwen3.8-Max-0902 cost?
At publication time, QwenCloud lists $2 per million input tokens and $6 per million output tokens, plus separate cache prices. Check the current service page because pricing can change or differ by region.
Is Qwen3.8-Max-0902 available in Chat AI?
Its availability is unverified. The exact model does not appear in Chat AI's current public model directory.
Evidence