AI/TLDR

Qwen3.8-Max-Preview

Preview of Alibaba's next Qwen-Max flagship — a 2.4-trillion-parameter multimodal MoE model unveiled at WAIC Shanghai on July 19, 2026.

Overview

Qwen3.8-Max-Preview is the preview of the next flagship in Alibaba's Qwen-Max line, unveiled by the Qwen team on July 19, 2026 during the World Artificial Intelligence Conference (WAIC) in Shanghai. Alibaba describes it as a 2.4-trillion-parameter model and as its first Qwen model above one trillion parameters to accept images, video, and documents alongside text — making it a multimodal successor to the text-only Qwen3.7-Max.

The preview is available through the Alibaba Cloud Token Plan subscription and the company's Qoder and QoderWork agentic platforms, priced during the trial period at 10% of the standard rate. Alibaba has not published a benchmark table, a model card, an activated-parameter count, or a formal license; the Qwen team has said the model is "continuously evolving" and that open weights will follow, without committing to a date.

Alibaba's own positioning claim is that Qwen3.8-Max-Preview is second only to Anthropic's Claude Fable 5 among today's frontier systems. No third-party leaderboard (Artificial Analysis, LMArena, or otherwise) had independent scores for the model at launch, so treat every quality claim about Qwen3.8-Max as unverified until the Qwen team publishes an official benchmark table or an external evaluator reports numbers.

Released2026-07-19
LicenseProprietary (preview); open-weight release promised by the Qwen team but no license or date announced.
WeightsAPI only
Parameters2.4 trillion total (active parameters not disclosed).
ModalitiesText, Image, Video, Documents
StatusPreview. Available through Alibaba Token Plan, Qoder, and QoderWork at 10% of standard pricing during the trial period. Open weights promised but no date announced.

Strengths

  • First Qwen model above one trillion parameters to accept images, video, and documents alongside text
  • 2.4-trillion-parameter scale, making it one of the largest models previewed to date (alongside Moonshot's Kimi K3)
  • Delivered as the flagship of the Qwen-Max agent line — positioned by Alibaba against Anthropic's Claude Fable 5
  • Preview access is available immediately through Alibaba's Token Plan, Qoder, and QoderWork at 10% of standard pricing

Best for

  • Frontier reasoning and general-purpose tasks where the previous Qwen-Max flagship was competitive
  • Multimodal analysis of images, video, and documents alongside text prompts
  • Coding, full-stack development, data analysis, and office workflows (the tasks the Qwen team called out in its preview announcement)
  • Note: no independent benchmarks exist yet — until they do, every performance claim about Qwen3.8-Max should be treated as unverified

Qwen-Max — every version

The full lineage of the Qwen-Max line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.

VersionReleasedContextLicense
Qwen3.8-Max-Preview2026-07-19Proprietary
Qwen3.7-Maxcurrent2026-05Proprietary
Qwen3-Max2025-09Proprietary
Qwen2.5-Max2025-01-29Proprietary

FAQ

Has Qwen3.8-Max been released?

It is available as a preview, not a general release. The Qwen team unveiled Qwen3.8-Max-Preview on July 19, 2026 at the World Artificial Intelligence Conference in Shanghai, offered through the Alibaba Cloud Token Plan, Qoder, and QoderWork at 10% of standard pricing during the trial period. Alibaba has not published an official model card, benchmark table, or license.

What are Qwen3.8-Max's specs?

Alibaba has disclosed only that Qwen3.8-Max has 2.4 trillion total parameters and is its first Qwen model above one trillion parameters to accept images, video, and documents alongside text input. The active-parameter count, the context window, the training data, and the architecture have not been officially published. Third-party recaps have suggested a 1M-token context window, but this is not confirmed by the Qwen team.

Is Qwen3.8-Max open-source or open-weight?

Not yet. Qwen3.8-Max-Preview is a proprietary preview served through Alibaba's own platforms. The Qwen team has said that open weights will follow, but has not published a license or a release date.

How does Qwen3.8-Max compare to Claude Fable 5 or Kimi K3?

Alibaba's own positioning claim is that Qwen3.8-Max is second only to Anthropic's Claude Fable 5 among the systems the Qwen team benchmarked. Independent leaderboards (Artificial Analysis, LMArena) had no scores at launch, so this ranking is Alibaba's internal claim, not a verified third-party result. Moonshot's Kimi K3, previewed three days earlier at 2.8 trillion parameters, is the other headline preview in this weight class.