Qwen3.8-LiveTranslate, a next-generation real-time translation large model, has been upgraded in translation quality, latency, speaker identification, and speech synthesis. Featuring an audio-text interleaved Interleave architecture, it reduces average latency from 2.8 to 2.3 seconds. Key capabilities include real-time speaker separation, bilingual synchronization, and long-context disambiguation. Using a Hybrid MoE Thinker-Talker design, it supports 60 languages and outperforms mainstream systems in fidelity and speaker separation. The API is available on the Qwen AI platform, with future focus on reducing latency and expanding language support.