Tencent HunyuanLaunchAI ModelsNovember 20, 2026

Tencent releases WeMM-Video omni-modal video models

Read original source →arxiv.org

WeMM-Video ships in four sizes — 2B, 4B, 9B, and 35B-A3B — as Chinese-English bilingual omni-modal models unifying text, image, audio, and video. It pairs hybrid thinking modes with a multi-stage post-training pipeline using multi-teacher OPD, anti-repetition MrDPO, and specialist RL.

2 sources

More stories today

Open the live feed