Tencent releases Hy4 multimodal MoE model with 1M-token context
The Tencent Hy Team introduced Hy4, a new-generation Mixture-of-Experts flagship model for text, image, and video understanding, according to its official model card. Its 770B-parameter language backbone activates 49B parameters per token and supports a 1M-token context, while a 3.3B DFly2 drafter is included for speculative decoding. Tencent also provides FP8 and GGUF artifacts, vLLM and SGLang deployment recipes, a finetuning pipeline, and documentation under the Tencent Hy License.
Hy4 combines multimodality, long-context processing, open deployment resources, and a large MoE architecture in a substantive flagship-model release, although its performance claims come from a vendor model card and have not been independently verified.
tencent/Hy4-FP8 Image-Text-to-Text • 774B • Updated about 3 hours ago • 2 • 1
license: other license_name: tencent-hy-license license_link: https://huggingface.co/tencent/Hy4-FP8/blob/main/LICENSE library_name: transformers pipeline_tag: image-text-to-text tags:
- hunyuan
- hy4
- moe
- multimodal
- instruct
- fp8
- quantized
中文 | English
#Tencent Hy4#Tencent Hy Team#DFly2#DeepSeek#GLM#vLLM#SGLang#AngelSlim