On April 2, the official
Core Power: Number One in Domestic Models, Close to Claude
In authoritative programming capability evaluations of the series,
Task Autonomy: In front-end web development and complex repository-level tasks, it can autonomously break down tasks, plan paths, and complete the "test-modify" loop, just like an experienced engineer.
Engineering Implementation: The model achieved the best results in long-term planning tasks, not only excelling at high-quality code repair, but also being proficient in handling complex terminal operations and automated task execution.
Visual Evolution: From "Understanding Design Drafts" to "One-click Web Generation"
Thanks to native multimodal data training,
Linkage Achieved: The model can directly generate front-end pages, complete code, and modify interactions based on interface screenshots, design drafts, or natural language descriptions.
Perceptive Sensitivity: It realizes a complete loop from "understanding the interface" to "using tools to make modifications," greatly shortening the development cycle for UI restoration.
Ecosystem Compatibility: One Million Context, Fully Compatible with Agent Frameworks
To unleash the maximum potential of intelligent agents, Qwen 3.6 has significantly enhanced its basic specifications:
Extended Vision: It supports a context window of up to 1 million tokens, easily capable of handling entire code repositories.
Full Framework Compatibility: It has optimized support for mainstream Agent frameworks such as
Industry Insight: Making "Atmosphere Programming" Truly Practical
The release of
Currently,




