千问办公首发上线Qwen3.8-Flash，生成速度提升100%，Token消耗减少75%

TL;DR - Qwen Office has launched a standard mode powered by an office-optimized Qwen3.8-Flash, claiming roughly twice the generation speed and 75% lower average token use on real office tasks. The release targets the typical trade-off between agent performance, latency, and cost.

• Qwen3.8-Flash is described as a new architecture with hundreds of billions of total parameters and performance exceeding Claude Opus 4.6, according to Qwen.
• The office-specific model was tuned for multi-step planning, tool selection, and context compression.
• Inference optimizations and a customized agent harness reportedly improve token throughput and reduce resource consumption.
• Qwen expects the standard mode to handle 95% of routine office tasks, reserving an advanced mode for the remaining complex workloads.