Liquid AI tunes Liquid Context for Snapdragon
Liquid AI has optimized Liquid Context, its on-device personal context layer, for Qualcomm's Hexagon NPU, so phone makers can add local user memory that any agent the user chooses can draw on.
Liquid AI announced at Snapdragon Summit 2026 that Liquid Context, its on-device context layer, is now optimized for Snapdragon processors, specifically the Qualcomm Hexagon NPU. The work was done in collaboration with Qualcomm Technologies.
With the user's permission, Liquid Context learns from device signals, models a user's routines, preferences and needs, and keeps that model up to date. According to Liquid AI, it runs in the background on the NPU and builds and maintains context locally, so a cloud model does not have to process every update. It then shares relevant context with connected agents according to the user's permissions. Those agents can run on the device, in the cloud or in a hybrid setup, and can include third-party agents as well as Liquid AI's own Liquid Agent.
Liquid AI has also optimized LFM2.5-2.6B, the agentic model behind Liquid Agent, for Hexagon. The company is offering Liquid Agent to OEMs as an embedded agent they can tailor to their own hardware, services and brand.
The examples Liquid AI gives are all illustrative: rescheduling meetings around a sick child, drafting a LinkedIn recap of a conference, and passing workout data from a smartwatch to a car's agent.
The announcement does not include performance figures, power measurements, named OEM customers or a shipping timeline. The two companies say they are "exploring additional opportunities" so OEMs can offer personal context as a built-in device feature. Qualcomm EVP Durga Malladi described the partnership as helping to "accelerate the next generation of Agentic AI experiences."