On-Device Models
Machine-learning models that run locally in software like Chrome — fast, private, and shaping how your content is read before it reaches a server.
On-device models are machine-learning models that run locally on your own hardware rather than in the cloud, giving fast responses and keeping data private. Chrome ships a small army of them, stored in numbered folders inside your user profile.
Their scope is broad. Chrome's on-device roster includes language detection, a text classifier for smart selection and entity extraction, a text embedder and passage embedder for semantic understanding, phrase segmentation, and text-safety models — plus named services like Orca (core text processing), Mahi (document intelligence and summarisation), and Walrus (content moderation).
For AI SEO this matters because content understanding increasingly starts in the browser, before anything reaches a search server. These models feed features like Chrome's history embeddings and reader-mode extraction via DomDistiller, making the browser itself a place where your content is read and judged.
Related concepts
Mentioned in
-
Chrome Decisions API: the prompts, limits, engines and code reviews behind DecisionModel
Chrome's DecisionModel API landed on 6 to 8 October 2026. The exact prompts, limits, engine status and review findings, plus a comparison with Google's MediaPipe DecisionMaker and local tests of EmbeddingGemma 1, EmbeddingGemma 2 and Laya.
