Liquid AI is pushing personalized artificial intelligence deeper into the device itself. At Snapdragon Summit 2026, the company announced that Liquid Context, its on-device context layer, has been optimized for Snapdragon processors and Qualcomm’s Hexagon NPU. The collaboration with Qualcomm Technologies is designed to give AI agents a persistent understanding of a user’s routines, preferences and needs without sending every piece of information to the cloud.
The announcement points toward a different kind of personal AI. Instead of forcing users to explain their situation every time they interact with an assistant, AI agents could already have access to relevant context, provided the user has given permission for that information to be used.
Liquid Context Gives AI Agents a Better Understanding of the User
Most AI assistants still depend heavily on prompts. Users often have to explain what they are doing, what they want and which details matter before the system can provide a useful response. Liquid Context is designed to reduce that friction by building an evolving picture of a user’s routines, preferences and current needs directly on the device.
With permission, the technology can process signals generated across a device and turn them into useful context for compatible AI agents. That information can then support agents running locally, in the cloud or through a hybrid setup. Rather than operating as another chatbot, Liquid Context works underneath AI agents as a shared context layer that helps them understand the situation before responding or taking an approved action.
Liquid AI says this personal context is built and maintained locally. Access to relevant information is also controlled through user permissions, which could become increasingly important as AI assistants gain access to more personal data.
Snapdragon’s Hexagon NPU Keeps More AI Processing on the Device
Running persistent AI context in the background could quickly become expensive and inefficient if every update had to be sent to a cloud data center. Liquid AI is addressing that challenge by optimizing Liquid Context for Qualcomm’s Hexagon NPU, allowing more of the processing to happen directly on Snapdragon-powered hardware.
The approach allows device signals to be converted into useful context without depending on a remote AI model for every interaction. That can reduce latency while also limiting how much personal information needs to leave the device.
Qualcomm has been moving in the same direction with its broader Snapdragon strategy. The company increasingly positions its chips as platforms for agentic AI that can understand personal context, coordinate tasks and operate across smartphones, PCs, vehicles and other connected devices. The Liquid AI collaboration fits neatly into that direction because it combines AI models with the hardware needed to run them continuously at the edge.
Personal AI Could Become More Useful Without Constant Prompting
Liquid AI outlined several scenarios that show how persistent context could change everyday AI interactions. One example involves a parent receiving a message that a child needs to leave school early. Instead of manually explaining their calendar, family situation and work priorities to an assistant, an AI agent could use approved context to identify meetings that may need to be moved and prepare rescheduling messages for review.
Another example involves someone returning from a conference. The context layer could retain useful information from the event alongside the user’s preferred writing style and selected photos. An AI agent could then use that information to prepare a LinkedIn recap without requiring the person to manually rebuild the context from scratch.
The company also sees opportunities across multiple devices. A smartwatch could capture activity during a run, while a connected vehicle could later use approved information from that activity to adjust recommendations or settings. These examples remain illustrative rather than confirmed consumer features, but they show the broader idea behind Liquid Context: AI that requires fewer explanations because it already understands more of the user’s immediate situation.
Liquid Agent Adds an Embedded AI Agent to the Platform
Liquid Context is not the only technology Liquid AI has optimized for Snapdragon hardware. The company has also optimized its LFM2.5-2.6B agentic model, which powers Liquid Agent, for Qualcomm’s Hexagon NPU.
Liquid Agent is designed as an efficient embedded AI agent that manufacturers can customize around their own devices, services, interfaces and brands. In this setup, Liquid Context maintains relevant user information and task state, while Liquid Agent can use that context to reason about what should happen next.
The architecture also gives manufacturers some flexibility. Liquid AI says the context layer can work with compatible third-party, cloud-based, embedded and hybrid AI agents. That means device makers would not necessarily need to rely on a single AI assistant to benefit from persistent personal context.
Qualcomm and Liquid AI See Personal Context as a Device-Level Feature
The collaboration could eventually allow device manufacturers to treat personal AI context as part of the operating experience rather than something tied to one specific chatbot or application. Liquid AI and Qualcomm are exploring ways for manufacturers to expose persistent context directly to users while connecting it with compatible AI agents and services.
That could give smartphone, PC and device manufacturers another route into the agentic AI market. Instead of simply installing a large language model on a device, companies could build AI systems that remember relevant context, understand ongoing tasks and preserve continuity between different applications and assistants.
For users, the difference could be noticeable. An AI system that already understands the current task, recent activity and personal preferences may feel far more useful than one that begins every interaction from zero.
On-Device AI Is Moving Beyond Simply Running Models Locally
For several years, much of the discussion around on-device AI focused on whether smartphones and laptops could run increasingly capable models without depending on cloud infrastructure. That challenge is still important, but the next stage of competition appears to be shifting toward context.
A locally running model can still feel generic if it knows very little about the person using it. Persistent context could help close that gap by giving AI systems access to relevant information about routines, preferences and ongoing tasks.
Liquid AI’s approach treats that context almost like infrastructure. The AI model can change, the device can change and the agent can change, while the underlying personal context remains available to approved systems. That approach could become especially valuable as users begin interacting with several specialized AI agents rather than relying on one universal assistant.
Privacy and Permissions Will Shape How Personal AI Develops
The benefits of persistent AI context also introduce an obvious privacy challenge. An assistant that understands schedules, communications, routines, preferences and activity can become far more useful, but it also handles information that users may consider highly personal.
Liquid AI says personal context is built and maintained locally, while connected agents receive relevant information according to user permissions. Processing on Qualcomm’s Hexagon NPU can also reduce the need to send every context update through a cloud service.
The real test will come from implementation. Users will need clear controls over what an AI agent can access, how information moves between devices and which actions an agent can perform. As personal AI becomes more capable, transparent permissions may become just as important as the intelligence of the model itself.
The Agentic AI Race Is Moving Closer to the Edge
Liquid AI’s collaboration with Qualcomm arrives as AI developers and chipmakers increasingly focus on moving intelligence closer to the user. Liquid AI describes itself as a device-native foundation model company focused on running advanced AI outside traditional data centers, particularly in environments where privacy, latency, memory and computing resources matter.
Qualcomm is taking a similar direction with Snapdragon. The company is increasingly framing its hardware around the rise of agentic AI, where assistants move beyond answering questions and begin understanding intent, coordinating tasks and taking permitted actions.
Liquid Context sits directly inside that shift. The important part of the announcement is not simply that another AI model can run on Snapdragon hardware. The bigger change is that the device itself could become the place where AI systems build and maintain the context needed to become genuinely personal.
That may turn out to be one of the most important parts of the next phase of on-device AI.
Sources
Business Wire — Liquid AI, in Collaboration with Qualcomm Technologies, Brings Personal AI Context to Devices Powered by Snapdragon
https://www.businesswire.com/news/home/20260923507264/en/Liquid-AI-in-Collaboration-with-Qualcomm-Technologies-Brings-Personal-AI-Context-to-Devices-Powered-by-Snapdragon
Qualcomm — Snapdragon for the Agentic Age
https://www.qualcomm.com/news/onq/2026/09/snapdragon-for-agentic-age
