Qwen office unveils Qwen3.8-Flash agent, enhancing office productivity with advanced AI capabilities

The Qwen Office Platform has officially launched the Qwen3.8-Flash large model, accompanied by a new standard mode, making it accessible to all users. This model is designed to streamline office tasks with reduced point consumption and improved Token throughput efficiency. The platform plans to establish a dual-model supply system, with the standard mode addressing 95% of routine office needs and the advanced mode handling the remaining 5% of complex tasks. The Qwen3.8-Flash model features a new architecture with hundreds of billions of parameters, achieving performance that surpasses Claude Opus4.6. A dedicated version of the model for office use has been developed through collaboration between the Qwen large model team and the Qwen Office team. This version is optimized for core Agent scenarios such as multi-step planning, tool calling, and context compression, with enhancements in reasoning optimization and a customized Harness architecture to boost system efficiency. Real-world office scenario tests indicate that the standard mode of Qwen Office has increased single-task generation speed by approximately 100%, while reducing average Token consumption by 75%. This advancement addresses a long-standing industry challenge in AI applications, where high performance typically correlates with high costs and latency. Qwen Office aims to break this 'impossible triangle' by integrating Agent technology with large models, enhancing cost-effectiveness and efficiency. As model intelligence continues to evolve and platforms optimize alongside underlying models, AI Agents are anticipated to overcome Token resource limitations, entering a new phase characterized by high efficiency and cost-effectiveness, often described as 'well-fed' in terms of resource availability.