What is the context window of Qwen4-Max-Preview?

Qwen4-Max-Preview features a 1-million-token input context window.

What is the maximum output limit for Qwen4-Max-Preview?

The model provides a maximum output limit of 393,216 tokens, which covers both internal reasoning and the final response.

What modalities does Qwen4-Max-Preview support?

It supports text, image, and video inputs.

Qwen4-Max-Preview: Features & Limits

Qwen4-Max-Preview API Specs and Pricing. A macro shot of a dense multi-chip module on a server motherboard, illuminated by a sharp amber LED edge-light, with crystalline condensation beading on the anodized black aluminum heatsink fins — GPUniq

Data as of November 12, 20263 sources

Drafted with AI tools and edited by a human. The figures and conclusions were checked by a GPUniq editor.

TL;DR

Qwen4-Max-Preview features a 1-million-token context window and a 393,216-token output limit. Access is restricted to the Alibaba Cloud Model Studio Token Plan during the preview period. The model supports text, image, and video inputs. It will be retired on November 25, 2026.

Context window capacity

This model handles 1,000,000 input tokens. This capacity lets you feed the model entire repositories or long technical manuals without splitting files.

Feature	Specification
Context Window	1,000,000 tokens
Max Output	393,216 tokens
Input Modality	Text, Image, Video
Release Date	Nov 11, 2026
Retirement Date	Nov 25, 2026

Maximum output tokens

The model provides a shared output budget of 393,216 tokens. This pool covers both the internal reasoning steps and the final visible response. It's built for deep logic. Use it when you need more than a simple summary.

API access requirements

Standard pay-as-you-go billing is unavailable during the preview. You must use the Alibaba Cloud Model Studio Token Plan. This subscription reduces credit burn. You'll find it in the QwenCloud console or through apps like QwenWork, Qoder, the Qwen App, and QwenStudio.

Preview period dates

The preview is brief. It launched on November 11, 2026. The model will be retired on November 25, 2026 (UTC+8). Migrate your production workflows before this cutoff to avoid service interruptions. Alibaba hasn't shared per-token pricing for use outside the Token Plan yet.

Supported modalities

This multimodal model processes text, images, and video. Use it for tasks requiring tool calling, web searches, or structured data output via JSON Schema. It is designed for complex reasoning across different media types.

API format compatibility

The API supports OpenAI Chat and Anthropic Messages formats. Switching existing codebases requires minimal effort.

{
  "model": "qwen4-max-preview",
  "messages": [
    {
      "role": "user",
      "content": "Analyze this codebase."
    }
  ],
  "max_tokens": 393216
}

The model remains proprietary. Alibaba has not released open weights, parameter counts, or specific benchmark results for this version yet. It serves as the successor to Qwen3.8-Max.

Methodology

Facts are compiled from the vendor's own announcement and documentation, linked below, at the time of writing; where the vendor hasn't published a number it is marked as such. GPUniq prices, when the model is already available here, come from our live catalog.