Back to feed
Alibaba Unveils Qwen4-Max-Preview with 1M Token Context

Alibaba Unveils Qwen4-Max-Preview with 1M Token Context

Alibaba Cloud has unveiled Qwen4-Max-Preview, a model designed for complex programming tasks and autonomous interaction. The model provides access to the new Qwen 4 family with enhanced capabilities for developers.

Technical Specifications

The new architecture supports a context window of up to 1 million tokens. The response generation limit has been increased to 393,216 tokens, allowing dynamic resource allocation between step-by-step analysis and final output.

Qwen4-Max-Preview is built as a multimodal system with native support for text, images, and video streams. This enables comprehensive analysis of heterogeneous data within a single interface.

Built-in Agent Mode

Agent Mode allows the model to independently decompose tasks and invoke external tools. The AI can execute code in an isolated environment, debug errors, and prototype web interfaces without human intervention.

The architecture was trained on real-world engineering scenarios, including deep code refactoring and vulnerability detection. This makes the tool effective for automating office processes and software development.

Availability and Integration

You can try the model for free via the Qwen Chat interface or through Alibaba Cloud's corporate Token Plan subscription. For professional development, integrations are available in Qoder and QwenWork environments, as well as the Model Studio platform.

5K views

More from this channel Metavershche and AIshche

Similar in this category Technology