AI infrastructure platform company Lablup will participate in CES 2026, to be held at the Las Vegas Convention Center in the United States from Jan. 6 to 9, for a fourth straight year.

At the event, Lablup will officially launch Backend.AI:GO, a desktop application that directly runs a small language model (SLM) on a personal PC without cloud servers.

In Backend.AI:GO, “GO” is an abbreviation for Generative On-device. It accelerates AI inference by using local resources in an on-premises environment.

After downloading the model, it works without an internet connection because it does not go through the cloud, and data entered by users is not sent to external servers. Users can easily download models on their personal computers and use them for internal document analysis, image understanding and generation, and code reviews without security concerns. Users can also adjust hyperparameters and resource allocation to change response characteristics and resource usage as desired.

Lablup also plans to officially launch its cloud-based product Backend.AI:DOL at CES.

Lablup CEO Jeong-gyu Shin said enterprise demand is rising quickly for running LLMs directly in on-device environments as the AI PC era gains momentum. He said Backend.AI:GO and DOL will provide an environment where AI infrastructure can be operated with a consistent experience regardless of scale and environment, from a single personal PC to GPU clusters of several thousand machines.