AI · SEMICONDUCTOR · FACT CHECK
SKD WIREMostly True
Liquid AI has released LFM3, a new on-device AI model capable of handling browser-based tasks, voice and image processing, moving its liquid foundation model lineup further toward local execution on end-user devices.
LFM3 is designed to run directly on devices rather than relying on cloud-based inference, according to Liquid AI. The model extends the company's liquid foundation model architecture, which the startup has positioned as an alternative to conventional transformer-based large language models (LLMs).
On-device AI has become an increasingly contested segment, with chipmakers and model developers alike seeking to bring inference workloads onto smartphones, PCs and edge hardware. Liquid AI's release of LFM3 places the company in that field with a model that spans multiple modalities.
The newly unveiled LFM3 supports browser-based processing alongside voice and image handling, broadening the range of tasks the model can perform locally. That combination — text-oriented language tasks together with audio and visual inputs — signals Liquid AI's intent to make LFM3 a general-purpose foundation model for device-level deployment rather than a single-purpose language model.
The company announced the model through its official channels. Specific benchmark results, parameter counts and supported hardware platforms were not detailed in the initial announcement, and details on licensing and availability remain limited pending further disclosure from Liquid AI.
Liquid AI, a spinoff rooted in research on liquid neural networks, has built its model family around architectures that differ from the dominant transformer design. The LFM series has previously drawn attention for its efficiency-focused approach, and LFM3 represents the next step in that lineup.
The move comes as the on-device AI market accelerates, with device makers and semiconductor suppliers pushing NPUs and other dedicated silicon to support local inference. Model releases such as LFM3 are aimed at giving that hardware workloads to run, and vice versa.
With browser, voice and image processing support on board, LFM3 rounds out Liquid AI's offering for developers building locally running AI applications. The claim that Liquid AI unveiled an on-device AI model named LFM3 supporting browser, voice and image processing is, on balance, Mostly True.
