vivo presents a new answer to its AI strategy
```
Author | Huang Yu
Editor | Wang Xiaojuan
In the past two years, focusing on AI smartphones has become a consensus among major smartphone manufacturers, but no winner has yet emerged. To become the “new king” of the AI era, smartphone makers now need to quickly “hone their internal skills” and create AI phones that truly revolutionize human-machine interaction.
As a leading player in the AI smartphone arena, on October 10 at the 2025 vivo Developer Conference, vivo presented the latest results of its AI strategy — the newly upgraded Blue Heart Intelligence strategy, clearly proposing the promotion of AI and operating system integration from three aspects: Blue Heart large model matrix, Blue Heart personal intelligence framework, and Blue Heart intelligence open platform.
A year ago, vivo officially released a new AI strategy — Blue Heart Intelligence, aiming to deeply integrate large model technology with smartphone operating systems to create personalized intelligence. vivo Senior Vice President and Chief Technology Officer Shi Yujian stated that under the AI wave, operating systems have now evolved fully to AIOS, which is an industry consensus and a necessity for users who demand more natural and intuitive experiences.
Now, what major smartphone makers are competing on is: who can better use AI to reshape the way humans interact with machines, so that AI phones can perceive more autonomously.
At the press conference, vivo vice president, OS product vice president, and head of vivo AI Global Research Institute Zhou Wei said that after a year’s efforts, vivo has steadily advanced the integration of AI and the operating system, gaining deep understanding of user preferences and scenarios, with a focus on enhancing Blue Heart large model matrix, Blue Heart personal intelligence framework, and Blue Heart intelligence open platform.
Large models are the base of AIOS’s capability. At the 2023 developer conference, vivo first launched its self-developed Blue Heart large model matrix. The Blue Heart large model matrix unveiled this time has achieved further breakthroughs in language, voice, and image, and specifically for personalized user needs, launched a lighter, more comprehensive 3B on-device multimodal inference large model.
On-device-cloud collaboration is the mainstream deployment scheme for AI large models in smartphones today. The on-device side handles AI tasks with lower computing power, such as translation, image editing, and local search, which has advantages in cost and privacy; while the cloud handles high-computing-power scenarios such as text-to-image, video editing, and online search, enabling more efficient use of resources and providing a smarter experience.
Zhou Wei said that in the past two years, vivo has achieved on-device deployment for multi-modal models of 13B, 7B, 3B, and 1B scales, and currently implemented 18 on-device functions on phones. “Based on practical experience with different model sizes, we have focused on the technical route of 3B model on-device deployment since last year.”
It is reported that through a year of continuous training, vivo developed the 3B on-device multimodal inference large model, which has capabilities in language, multimodality, logical reasoning, 128K long context, and UI Agent.
Zhou Wei stated that the UI Agent capability enables the model to understand phone interfaces and operate the phone. vivo has built specially designed training data for mobile UI operations, allowing the model to understand the phone’s interface and execute cross-application operations. Therefore, the 3B on-device multimodal inference large model is also the world’s first 3B model specifically built for on-device Agents.
Of course, for AI phones to sense more autonomously and better understand user intentions, not only is an improvement in large model capability required, but the large model also needs better integration with the system.
It is reported that vivo’s new Blue Heart personal intelligence framework not only allows large models to precisely understand user intentions through global sensing and multimodal data fusion, but also deepens understanding of the user through ongoing personalized data accumulation.
At the same time, Zhou Wei pointed out, this technology also lays the foundation for vivo to develop AI glasses and future embodied intelligent brains.
With the upgraded Blue Heart large model matrix and the brand new Blue Heart personal intelligence framework, vivo can further integrate AI and the operating system. However, for AI phones to become true AI Agents, in addition to enhanced autonomous perception, it is also necessary to enable access to third-party apps, break data silos of individual apps, and build an AI ecosystem that connects with third-party apps.
To achieve this goal, there are currently two technical routes in the market: First is the intent framework, which requires third-party app authorization but is relatively mature technologically and demands low hardware computing power; second is the pure visual solution, which does not require authorization, but its technology still needs improvement and it requires higher hardware computing power.
At present, vivo is more focused on the intent framework strategy. At this conference, it also announced a comprehensive upgrade of the Blue Heart Intelligence Open Platform, including platform, protocol, and distribution layers, opening up Blue Heart Intelligence's personalized capabilities to more developers.
Last year, vivo released the Intent Framework White Paper. This year, Intent Framework has been upgraded to 2.0, with full MCP protocol compatibility. At the same time, vivo launched the adaptation agent protocol A2A: "With this, developers can create agents and configure cards entirely online based on the Agent protocol, and then distribute them across all vivo channels via the unified intent framework and agent container," Zhou Wei said.
The next-generation mobile operating system OriginOS 6 is vivo’s new achievement in reconstructing the operating system with AI large models.
A major feature of OriginOS 6 is that, based on on-device “personal exclusive models,” local large models understand users’ needs and scenarios, providing fully personalized services through learned preferences. “Upgraded multimodal interaction will also proactively bring services to users.”
From an AI feature perspective, OriginOS 6 fully upgrades “Xiao V Circle Search,” supporting automatic identification of the main content on-screen and intent understanding. It moves from “search and recognize” directly to “instant execution,” enabling one-step access to specific app services; the built-in Xiao V Memory 2.0 can use AI to distill and organize saved content, and also connect schedule and address information.
Additionally, the Blue Heart Xiao V intelligent voice assistant can call over a thousand mobile and ecosystem application skills.
Apart from the Android-based OriginOS 6, vivo continues to focus on self-developed operating systems and launched Blue River Operating System 3. At the 2023 developer conference, vivo launched the Blue River Operating System for the first time, the industry's first full-stack operating system written in Rust language, which debuted on watches.
Zhou Wei stated that Blue River Operating System 3 (BlueOS 3) is designed for native AI devices. For AI-native smart devices, BlueOS 3 builds a “stereoscopic perception system.” It improves camera startup speed and supports visual recognition and spatial algorithms.
In the future, vivo may equip more native AI devices such as smart glasses with Blue River Operating System.
As the AI era accelerates, vivo’s AI strategy has become increasingly clear. The next step is to see whether it can achieve further breakthroughs.
Risk Warning and DisclaimerThe market has risks, and investments need to be made cautiously. This article does not constitute individual investment advice, nor does it take into account the particular investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situation. Investing based on this information is at your own risk. ```