JD open-sources full-stack JoyAI-VL-Interaction, moving from Q&A to watching while talking
Original title:京东全栈开源JoyAI-VL-Interaction,让大模型从“一问一答”走向“边看边说”
AI summary
JD recently open-sourced JoyAI-VL-Interaction, the world's first full-stack interaction model, with native support from vLLM-Omni. The model can continuously observe video streams, proactively identify key events and respond in real time, and supports delegating complex tasks to a background agent. In 58 blind evaluations by real people, it achieved a 77.6% win rate against the Doubao video call assistant and an 87.9% win rate against the Gemini video call assistant, with a 100% win rate in monitoring and alert scenarios. The open-source release includes model weights, interaction datasets, training plans and a complete deployable system, supporting video inputs such as cameras and live streams as well as voice interaction, long-term memory and vLLM deployment, suitable for real-time scenarios such as security monitoring, elderly care and live-stream commentary.
Key Takeaways
- JD recently open-sourced JoyAI-VL-Interaction,the world's first full-stack interaction model,with…
- The model can continuously observe video streams,proactively identify key events and respond in…
- In 58 blind evaluations by real people,it achieved a 77.
- 6% win rate against the Doubao video call assistant and an 87.
- 9% win rate against the Gemini video call assistant,with a 100% win rate in monitoring and alert…
Quick Facts
- Published
- Updated
- Source
- 公众号:京东JoyAI
- Category
- Model Releases
- AI score
- 73
京东这个实时交互模型不是又一个刷榜模型,它把「什么时候该说话」变成了模型自己的判断。全栈开源且对比豆包、Gemini 胜率惊人,做安防、看护、直播的开发者可以直接试了。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: JD open-sources full-stack JoyAI-VL-Interaction, moving from Q&A to watching while talking. June 23, 2026. https://new.nat6.net/en/news/cmqq8ww0e07m4slp5mk5snl0z/