88Z AI, 15 July — Xiaomi has open-sourced Xiaomi-Robotics-U0, a 38-billion-parameter multimodal autoregressive embodied generative foundation model that unifies four task types including scene generation, with inference efficiency improved by nearly 83 times.
At a glance
- First embodied generative model unifying scene generation, trajectory transfer, interactive video, and image editing
- Equipped with FlashAR+ acceleration, improving inference efficiency by nearly 83 times
The update
Xiaomi has officially open-sourced Xiaomi-Robotics-U0, a multimodal autoregressive embodied generative foundation model with 38 billion parameters. As the first unified generative model in the embodied intelligence field, it simultaneously handles four core tasks: scene generation, trajectory transfer, interactive video, and general image editing. The model employs five-dimensional decoupled structured control technology to achieve precise and controllable content generation. Furthermore, by integrating the FlashAR+ inference acceleration solution, the model's operational efficiency has been significantly optimized, with overall efficiency improved by nearly 83 times, providing efficient infrastructure support for embodied intelligence applications.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 小米技术Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.