ModelsMulti-sourceSelected

vLLM 首日适配腾讯 Hunyuan Hy4 多模态模型

Why we picked it

腾讯把文本、图像、视频塞进一个模型,vLLM 发布当天就能跑。做文档理解或视频分析的工程师可以省掉拼多个模型的麻烦,部署直接上 vLLM 即可。

腾讯 Hunyuan 发布 Hy4 模型,开源推理框架 vLLM 在发布当天即提供支持(Day-0 support),开发者可立即用 vLLM 部署该模型。

Body

Hy4 是单一模型同时处理文本、图像和视频三种输入:可以阅读视觉化文档(如 PDF、图表截图)、对多张图片进行对比分析,以及对视频片段进行推理。这种三模态统一在一个权重里的设计省去了用户为不同任务切换多个模型的成本。

对开发者而言,Day-0 适配意味着无需等待社区移植,即可在生产环境中跑起该模型。vLLM 此前已对 Qwen、DeepSeek 等中国团队的开源模型提供早期支持,本土旗舰模型与主流推理框架的衔接速度正在加快。

Original · vLLM

🎉 Day-0 support for @TencentHunyuan’s Hy4 in vLLM!

Text, images and video in one model: read visual documents, compare images and reason over clips alongside text.

1/4 https://t.co/vftPPimxj7

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Hunyuan, vLLM, Tencent, multimodal.

Source attribution: curated from vLLM (@vllm_project). Cross-source corroboration: 4 independent sources have covered this development (including Together AI, Hunyuan, The Straits Times: Business), confirming industry-wide signal strength.