Tencent Hunyuan Open-Sources Hy-SimulMT Simultaneous Interpretation Model
Tencent Hunyuan released and open-sourced Hy-SimulMT, a 30B-A3B simultaneous interpretation model under the Apache-2.0 license with inference and training tools on GitHub…
Tencent Hunyuan released and open-sourced Hy-SimulMT, a 30B-A3B simultaneous interpretation model under the Apache-2.0 license with inference and training tools on GitHub, the Daily AI Brief reported October 26. The model supports real-time Chinese-English speech-to-speech translation with voice cloning, faithfully reproducing the speaker's timbre and tone. On the RealSI long-audio benchmark its StreamLAAL latency is about 3 seconds, the lowest among evaluated systems, and its GEMBA translation quality beats the compared open models and four of five commercial APIs. It is already available free on Tencent's Hy translation app.
Source: Daily AI Brief / Tencent, Oct 26, 2026 | 2026-10-26
延伸阅读：
- Nanjing Launches 5-Billion-Yuan Physical AI and World Model Fund
- China's First Open-Source Tax Large Model 'Lingjing' Released on ModelScope
- Qiyuan Robotics Ships 5,000th Personal Robot, Daily Output Past 200 Units
- AgiBot Unveils G2 Max Safety-Grade Heavy-Duty Embodied Robot With 40kg Dual-Arm Load
- Chinese Models' Weekly Token Calls Top U.S. for 26 Straight Weeks; DeepSeek Leads
- Tencent Hunyuan Open-Sources Hy-SimulMT Simultaneous Interpretation Model
- XPeng Wins Guangzhou Citywide L4 Robotaxi Demonstration Operating License
- Yiwu's 'Three Choices': From Price Wars to a 800-Billion-Yuan World Supermarket
- China AI Ambition on Show as DeepSeek Raises 100 Billion Yuan, Open-Sources DeepInfer
- Moonshot AI Releases Kimi K3.1 Pro: 3-Trillion-Parameter World's Largest Open-Source Model
