On June 2, 2026, Alibaba Tongyi Laboratory was officially releasedQwen3.7-PlusThe multi-modal agent model deeply integrates visual capabilities based on the text and agent capabilities of Qwen3.7 to realize an end-to-end closed loop of GUI operations, CLI calls, code generation and self-verification.
core competencies
The biggest breakthrough of Qwen3.7-Plus lies in the deep integration of "multi-modal + intelligent agent".Operation graphical interface (GUI)- Click buttons, fill out forms, drag elements - interact with the software like a real user.Command line (CLI)Operate execution code, manage the system, and deploy applications.
- multimodal understanding: Integrate visual capabilities to understand interface screenshots, analyze charts, and identify the location of UI elements.
- GUI Agent: Directly operate desktop applications and web pages, automating tasks without the need for APIs
- CLI Agent: Execute code, install dependencies, and manage file systems through the command line
- Code generation and self-verification: Automatically run tests and fix errors after writing code, forming a closed loop of development
- Open source: Continuing the open source tradition of Qwen series, developers can freely deploy and fine-tune
User experience and limitations
The 11-hour independent APP development demonstration at the press conference was impressive - from understanding the requirements to designing the interface, writing code, debugging and repairing, and final deployment, the entire process did not require manual intervention.
However, there are still some limitations in actual use: the stability of GUI operations is greatly affected by interface changes, and the planning and execution of complex tasks sometimes go "awry"; CLI operations are not friendly enough to novice users; as a new model, ecology and community support still need time to grow.
Overall Score
| 维度 | Score | evaluate |
|---|---|---|
| functional completeness | 8.8 / 10 | Multi-modal + GUI + CLI + code, the agent has comprehensive capabilities, but the ecosystem has just started. |
| 易用性 | 8.0 / 10 | There are thresholds for CLI operation, GUI stability needs to be improved, and documentation still needs to be improved. |
| Cost-effectiveness | 9.2 / 10 | Open source is free and can be deployed locally. It has significant cost advantages compared to closed source solutions. |
| 中文支持 | 9.5 / 10 | Chinese native model, Vision Arena ranks first in China, and has the best Chinese understanding |
| 输出质量 | 8.5 / 10 | Code generation and Agent execution are of high quality, with occasional deviations in complex tasks. |
Overall rating: 8.8/10
Qwen3.7-Plus is one of the most groundbreaking open source Agent models in 2026.
For more AI model reviews and tool recommendations, please visit AI Dash ——Discover the best AI tools.