Alibaba releases the new generation model Qwen3.6-Plus, with programming performance close to the world's strongest programming model

Alibaba releases the new generation large language model Qwen3.6-Plus.
Qwen 3.6 possesses native multimodal understanding and reasoning capabilities, with significantly enhanced overall performance.
In authoritative evaluations such as the SWE-bench series for intelligent agent programming and real-world agent tasks like Claw-Eval, Qwen 3.6’s programming performance approaches that of the world’s strongest programming models, the Claude series.
In practical scenarios such as front-end web development and complex repository-level tasks, Qwen 3.6 can autonomously decompose tasks, plan paths, test and modify until the task is completed, demonstrating a new breakthrough in multimodal agentic coding.
This makes “atmospheric programming,” where AI writes code driven by a single sentence, feasible.
Currently, Qwen 3.6-Plus is available on Alibaba Cloud Bailing, with a minimum cost of 2 yuan per million tokens input;
Qwen 3.6 has also been launched on Alibaba AI applications and platforms such as Wukong and Qwen APP.
(Science and Technology Innovation Board Daily)