Alibaba Qwen Releases Qwen3.8-LiveTranslate, Lowering Latency to 2.3 Seconds Across 60 Languages

Alibaba Qwen has released its next-generation real-time simultaneous interpreting model, Qwen3.8-LiveTranslate, based on the Interleave architecture, reducing the average latency (LAAL) from 2.8 seconds to 2.3 seconds across 60 languages. According to official information, the new model also demonstrates improvements in fidelity, fluency, and conciseness. This release marks another iteration of Qwen for simultaneous interpreting scenarios, with the latency data being the primary change.