Alibaba Revolutionizes AI with Autonomous Qwen3.7-Plus

Alibaba recently unveiled Qwen3.7-Plus, an artificial intelligence model that combines visual understanding capabilities with those of an autonomous agent. This model is designed to operate independently on graphical interfaces and applications, offering a new approach to automating complex tasks.

During testing, Qwen3.7-Plus demonstrated its ability to recreate desktop applications and program autonomously. A notable example is the creation of an English vocabulary learning application, where the model generated over 10,000 lines of code in eleven hours, thanks to more than 1,000 agent calls. This process included documenting requirements, generating code, installation, and version management.

In another demonstration, the model recreated the Stocks application from macOS, generating SwiftUI code and integrating an API for real-time stock data. Qwen3.7-Plus also showcased its capabilities in cloud task management through a browser extension, Qwen for Chrome, automating the purchase and configuration of virtual servers.

However, despite its impressive performance in navigating graphical interfaces, Qwen3.7-Plus lags behind in pure logic benchmarks compared to models like Claude Opus 4.6. In multimodal reasoning tasks, its results are mixed, outperforming some visual tests but failing on more complex scientific tasks.

Qwen3.7-Plus stands out for its compatibility with multiple frameworks, including the Anthropic API protocol, and works with Claude Code and OpenClaw. It offers a unique feature, preserve_thinking, which retains the reasoning from previous conversations, recommended for agent tasks.

Available through Alibaba Cloud Model Studio, Qwen3.7-Plus is offered at a competitive rate, costing $0.40 per million input tokens and $2.40 per million output tokens, making it significantly more affordable than leading Western models.