Samsung builds its own generative AI model, Gauss

Remember the dedicated Bixby button on the Galaxy S9 that everyone tried to remap? Samsung is back with another in-house AI push, but this time it actually makes technical sense.

Samsung Research officially revealed its own generative AI model, dubbed Samsung Gauss. Named after mathematician Carl Friedrich Gauss, who established normal distribution theory in the 1800s, the model is designed to power on-device features across Samsung hardware.

Three models for on-device tasks

Gauss is split into three main components rather than a single monolith.

Gauss Language takes care of tasks like summarizing documents, writing emails, and handling language translations.

Gauss Code acts as an in-house coding assistant to streamline software development.

Gauss Image handles prompt-based image generation, style modifications, image expansion, and upscaling low-resolution photos directly on the device.

The timing lines up with Qualcomm’s new Snapdragon 8 Gen 3 chipset, which supports on-device generative AI operations and is widely expected to power next year's Galaxy S24 lineup. Samsung is already teasing upcoming Galaxy AI capabilities, including real-time translation.

Why it matters

For smartphone users, the appeal here isn't beating GPT-4 in raw benchmark trivia. It is about local execution. Running these models directly on your phone means features work without an internet connection, without latency lag, and without sending private documents or photos to cloud servers.

While it is easy to be cynical given Samsung's past attempts to force Bixby onto users, this pivot looks much smarter. Cloud-based AI assistants like ChatGPT are expensive to run and useless offline. By combining in-house models with next-gen silicon on the upcoming Galaxy S24, Samsung is forcing the entire smartphone industry into an on-device AI race where privacy and offline utility matter more than parameter counts.