Qwen3.8 Max Preview

Qwen3.8 Max Preview sits within Alibaba's Qwen family of large language models and was published as a pre-release checkpoint of the Qwen3.8 Max line. A third-party benchmark tracker attributes the model to Alibaba and explicitly marks the Preview slug as superseded by the general-availability Qwen3.8 Max release, indicating that developers evaluating the Preview should expect a newer sibling to be the current production target in the same lineage.

The Preview variant is documented only through third-party aggregator pages rather than an official Alibaba model card, which limits the technical detail available for this specific checkpoint. The available evidence points to a vision-language design and a large context scope consistent with the wider Qwen3.8 generation, but architectural specifics such as parameter count and training methodology are not directly tied to the Preview slug in the supplied sources. Practical fit therefore centers on teams already invested in the Qwen ecosystem who want to compare a Preview snapshot against the current Qwen3.8 Max release before committing production traffic.

AIHubMixqwen3.8-max-previewqwen

Open provider page Provider docs

Quick Info

Provider: AIHubMix
Model key: qwen3.8-max-preview
Release date: Jul 19, 2026
Last updated: Jul 19, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.338
Output token cost: $1.014

Limits

Output tokens: 131,072 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Qwen3.8 Max Preview pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.8 Max Preview

AIHubMix

CoveragePreview

qwen3.8-max-preview is a hosted preview on Qwen Cloud's Token Plan, with the exact model ID confirmed by the Qwen team and a 1M-token context window per the Qwen Cloud model table. The vendor-reported total is 2.4T parameters, though architecture and active parameters are not publicly documented, and no downloadable weights exist as of the verification date. The July 20, 2026 follow-up build emphasized gains in web frontend work, but those claims remain vendor-reported without a public benchmark table or reproducible evaluation. Thinking, function calling, and built-in tools are listed as supported; structured output is not yet listed. A formal Qwen3.8 open-weight release remains planned but unscheduled.

AIHubMix

CoveragePreview

On July 19, 2026, Alibaba's Qwen team previewed Qwen3.8-Max-Preview, described as a 2.4 trillion-parameter multimodal model and Alibaba's first native trillion-plus-parameter system. Access runs through Alibaba's Token Plan subscription at 10% of standard pricing, with the preview offered as a purchasable hosted model during the WAIC conference in Shanghai. The Qwen account framed it as comparable to leading frontier systems and announced an upcoming open-weight release, though no benchmark table, model card, or license accompanied the announcement. The launch arrived two days after Moonshot AI released Kimi K3 as a 2.8T-parameter open-weight competitor, framing the competitive context.