Wukong is the first to integrate the country's most powerful programming model Qwen3.6

ME News update, April 2 (UTC+8). On April 2, Alibaba officially released its next-generation large language model, Qwen3.6-Plus. Alibaba’s flagship enterprise AI application, Wukong, was the first to complete integration.

Qwen3.6-Plus has seen major overall performance improvements across capabilities such as code, agents, reasoning, and native multimodal. In authoritative benchmark evaluations including the SWE-bench series for agent programming and the real-world agent task Claw-Eval, Qwen 3.6’s programming performance outperforms GLM-5 and Kimi-K2.5—models with 2 to 3 times more parameters—making it the domestic model with the strongest current programming ability, approaching the globally strongest programming model, the Claude series.

For Wukong users, using the new model brings three major improvements: agent programming capability is significantly enhanced compared with other general-purpose models; AI advances from a “co-pilot” to a “collaborator” that can independently take on sub-tasks, including the ability to write cross-file code, run tests, and iteratively fix issues on its own; long-range task planning and execution capabilities are substantially improved—when faced with the multi-step complex workflows common in enterprise scenarios, it can more reliably break down objectives, allocate resources, and deliver results. At the same time, its cost-performance advantage is prominent: the input cost is as low as 2 yuan per one million Tokens, greatly lowering the barrier for enterprises to adopt it at scale.

As Alibaba’s flagship enterprise AI application, Wukong plays a critical role in bridging the latest model technology advances with real enterprise needs. It has already been the first to integrate Qwen3.6-Plus, delivering an enterprise-grade Agent usage experience with stronger programming capabilities, better execution outcomes, and higher cost performance.

For example, in R&D scenarios such as “one-person development,” users can describe requirements in natural language, and Wukong independently completes the end-to-end process—from方案 breakdown, code writing, website generation, to test validation. In knowledge-management scenarios such as “one-person legal,” Wukong can extract key information across documents, compare changes in clauses, identify risks, and generate decision recommendations. In business process automation scenarios such as finance and manufacturing, employees only need to state the goal, and Wukong can independently carry out operations across systems, such as querying, filling, and approvals—turning tedious workflows into an intelligent service “in one sentence.”

Wukong’s early integration of Qwen3.6-Plus is a key practice of combining Alibaba’s AI models with real scenarios.

A product负责人 of Alibaba Wukong said that the competition for enterprise AI applications, in essence, is the product of “model capability × scenario understanding.” The breakthroughs of Qwen3.6-Plus in agent programming and long-range task planning provide a new foundation for Wukong to address complex pain points for enterprise customers. Meanwhile, the scenario understanding accumulated over many years of deep investment in the enterprise market by DingTalk helps Wukong know how to precisely embed these capabilities into real work workflows.

Starting immediately, Wukong membership users on v0.9.20 and later versions can get early access to the Qwen3.6-Plus model.

(Source: Ifnar)