Apple’s StreamBridge video model framework is released, Meta/WiMi accelerates the innovation boundary of multimodal AI technology!
On May 13, a technology media published a blog post, reporting that Apple (AAPL) and Fudan University jointly launched the StreamBridge end-side video large language model (Video-LLMs) framework to help AI understand live streaming videos.
Generally speaking, traditional video large language models are good at processing static videos, but cannot adapt to scenarios that require real-time perception such as robotics and autonomous driving. In these scenarios, the model is required to quickly understand the content of live video streams and respond.
Developing StreamBridge framework and innovative technologies
To solve the above problems, Apple and researchers from Fudan University developed the StreamBridge framework. The framework was tested on mainstream offline models such as LLaVA-OV-7B, Qwen2-VL-7B and Oryx-1.5-7B.
In addition, the research team also launched the Stream-IT dataset, which contains about 600,000 samples, combines video and text sequences, supports a variety of instruction formats, and aims to improve streaming video understanding capabilities.
Meta develops new AI model applications
At the same time, it is also worth noting that Meta (META) launched the “Meta AI” APP, which is strongly bundled with Ray-Ban Meta for the first time. Meta AI is an AI assistant created by Meta, driven by its own Llama large language model.
The latest data shows that Meta AI’s monthly active users are close to 1 billion, and the main entry point for users is naturally the social applications mentioned above to experience related functions. At the end of last month, Meta held its first AI developer conference LlamaCon, during which Meta launched its latest Llama 4 series of large language models.
It is undeniable that big models continue to empower industry development, the AI digitalization wave is surging forward, accelerating the pace of transformation and development in various industries, and various companies are actively carrying out AI scene construction, successfully completing the local deployment and scene adaptation of big AI models, applying AI to actual business scenarios, and promoting the development and upgrading of AI technology.
Wimi Hologram Cloud Inc deploys AI ecology to expand the boundaries of innovation
In the surging global technological wave, big AI models are reshaping the world at an unprecedented speed. In this process, data shows that WiMi (WIMI), as an innovative representative in the field of AI, has carried out in-depth layout around open source ecology, multimodal technology, computing power infrastructure and vertical scene applications, constantly breaking through the boundaries of AI technology and broadening the industry ecology.
From the introduction, WiMi builds a “holographic cloud” platform covering the cloud and edge through open model code, computing power interface and technical tool chain, supporting developers to call general big models such as DeepSeek for secondary development, and accelerate the commercial verification of vertical model applications.
At the same time, WiMi accelerates the landing speed of large models in application scenarios. The company has successively disclosed its more mature AI ecological landscape, covering the automotive, smart terminal, Internet, finance, education and scientific research, retail consumption and other industries, injecting strong momentum into the application of AI large models, and quietly becoming the key “fuel tank” behind this large model transformation.
Conclusion
As a transformative technology, the large model technology of artificial intelligence breeds “great development”. One of its important breakthroughs is to show “emergence ability” – when the model parameters continue to accumulate to the order of 10b (b represents the order of one billion), its performance (such as general knowledge ability, scientific reasoning ability, generation ability, etc.) shows nonlinear growth. So, we might as well look forward to more influential and empowering large models in thousands of industries in the market, stimulating industry momentum and industrial potential.
