Qwen announced the upgrade of Qwen3-Coder and the launch of the qwen3-coder-plus interface in Alibaba Cloud Model Studio. This new version enhances terminal task capabilities for practical engineering applications (including improved Terminal Bench metrics), achieving a public SWE-Bench score of 69.6 . It also introduces more secure code generation and constraint mechanisms. Officials claim that this model is compatible with Qwen Code and Claude Code workflows, making it suitable for scenarios such as long-term remediation, code review, and tool collaboration.
At the same time, Qwen Code has added multimodality and sub-agents , allowing for input of images (such as screenshots and error messages) during collaboration. Sub-agents then perform retrieval, modification, and testing, improving cross-file and cross-toolchain closed-loop efficiency. The API documentation and model pages have been updated, and developers can activate and integrate them in the console using examples.
Frequently Asked Questions
Q: What is the difference between qwen3-coder-plus and previous versions?
A: It performs better in terminal tasks and long processes, has improved SWE-Bench indicators, and has added stricter security and compliance policies.
Q: Can it be used together with Claude Code and Qwen Code?
A: Yes. The model is designed for proxy encoding and supports integration with both and terminal/tool calls.
Q: What is the use of Qwen Code's multimodality?
A: You can directly upload screenshots or interface images and combine them with text descriptions to locate the problem. The sub-agents will divide the work to complete the modification and testing.
Q: Where is qwen3-coder-plus called?
A: Open the model page in Alibaba Cloud Model Studio and access it according to the API example; online testing and usage viewing are supported.
Q: How do you understand the SWE-Bench score of 69.6?
A: This is one of the highest values in public evaluations, reflecting the comprehensive ability in real-world repair tasks. The specific settings are subject to the evaluation and documentation.
