Qwen3.8-Max
Quick Answer
Qwen3.8-Max is Alibaba’s newest flagship model, previewed on July 19, 2026 at the World AI Conference in Shanghai. Alibaba says it has 2.4 trillion total parameters, uses a sparse Mixture-of-Experts architecture, is multimodal (text, images, video, documents), and supports a 1M-token context window.
What to Watch
This is a preview release, not a finished, fully documented launch. Alibaba has not published a model card, an active-parameter count, or benchmark scores, which is a real gap compared to the company’s own prior Qwen3.7-Max release in May 2026, which did include that detail. Alibaba has also said an open-weight version is coming, without confirming a date, license, or repository.
Until those specifics are public, treat Qwen3.8-Max as promising but unverified. This page will be updated once Alibaba publishes fuller documentation.
What Qwen3.8-Max Is Best For (Based on Current Preview)
- Early experimentation: testing Alibaba’s newest model ahead of a fuller release
- Multimodal tasks: text, image, video, and document input, per Alibaba’s stated scope
- Long-context work: a 1M-token context window, if it performs as described
- Watching the open-model landscape: a signal of where Alibaba’s Qwen line is heading
Key Facts (What’s Confirmed vs. Not)
Confirmed by Alibaba:
- Developer: Alibaba Cloud
- Previewed: July 19, 2026, at WAIC 2026 in Shanghai
- Type: Multimodal, sparse Mixture-of-Experts model
- Total parameters: 2.4 trillion (stated by Alibaba)
- Context window: 1M tokens (stated by Alibaba)
- Current access: Preview only, through Alibaba Cloud’s Token Plan
Not yet published by Alibaba:
- A model card
- Active-parameter count
- Benchmark scores
- Open-weight license, repository, or release date
How to Access It
Preview access: through Alibaba Cloud’s Token Plan, aimed at developers who want to experiment early.
Open-weight release: not yet available. Alibaba has said this is coming, but nothing is confirmed. Check Alibaba’s official Qwen channels before planning around a self-hosted deployment.
Honest Limitations
- Unusually undocumented for a flagship launch: no model card or benchmarks as of this writing
- “Second only to Fable 5” is an unverified claim: made without supporting data
- Preview status: features, pricing, and access may change before a full release
- No confirmed open-weight terms yet: cannot be evaluated for self-hosting until Alibaba publishes a license
Alternatives Worth Knowing
- GLM 5.2, an already fully documented open-weight model from Z.ai with published benchmarks
- Kimi K3, Moonshot AI’s newly released open-weight frontier model
- DeepSeek, a widely used open-leaning model with a track record of published specs
- Claude and ChatGPT, closed frontier alternatives with full documentation
For how to weigh a model like this against fully documented options, see Open Models vs Closed Models.
Continue learning
Explore related guides, tools, workflows, and prompts that help you go deeper into this topic.
See how this tool fits into a workflow
Browse step-by-step AI workflows that use ChatGPT, Claude, Gemini, and other tools.
Frequently Asked Questions
Is Qwen3.8-Max fully released?
No, not yet. Alibaba unveiled Qwen3.8-Max Preview at WAIC 2026 in Shanghai on July 19, 2026. This is an early-access preview available through Alibaba Cloud's Token Plan, not a general public or open-weight release. Alibaba has said open weights are coming but has not given a date, license, or repository.
What is Qwen3.8-Max's parameter count?
Alibaba states the model has 2.4 trillion total parameters using a sparse Mixture-of-Experts architecture. Unlike the prior Qwen3.7-Max launch, Alibaba has not published the active-parameter count, a model card, or benchmark scores for Qwen3.8-Max, which is an unusual gap for a flagship announcement.
How does Qwen3.8-Max compare to other frontier models?
Alibaba has described it as second only to Claude Fable 5, but this claim was made without accompanying benchmark data or independent verification. Treat it as an unverified marketing claim until Alibaba or a third party publishes supporting evidence.
Should I build on Qwen3.8-Max right now?
For production or business-critical use, it's reasonable to wait. The preview is fine for experimentation, but without a published model card, benchmark scores, or a confirmed open-weight license, it's hard to evaluate fit or plan a long-term deployment around it yet.
Last updated: