Chinese internet giant Tencent launched its Hunyuan large model yesterday, featuring 13 billion parameters, and open-sourced its video-generation capabilities, supporting both Chinese and English input. Available on the Tencent Yuanbao app, users can apply for trial access in the AI Video section. Enterprise clients can integrate the model through Tencent Cloud, with API access currently open for internal testing. According to Tencent, Hunyuan is built on the DiT architecture with a number of upgrades, allowing developers and businesses to leverage pre-trained capabilities and thus save on resources. Tencent has previously open-sourced other Hunyuan models, including those for text-to-image and 3D generation. [IThome, in Chinese]

Related