General · EdgepediaSeptember 17, 202610 min read

Galaxy AI

Galaxy AI is Samsung Electronics' umbrella brand for a bundle of artificial-intelligence features on Galaxy phones, tablets, watches, and earbuds, launched with the Galaxy S24 series in January 2024. It is not a single model or app: a Samsung official has described it as a "unified AI brand" that blends Google's Gemini with Samsung's own software and device integration¹, and it combines Samsung's on-device processing with cloud services rather than running as one chatbot².

Key fact	Detail
What it is	A collection of AI features across Samsung apps and hardware, not a single app or chatbot²
Launch	Debuted with Galaxy S24 Ultra, S24+, and S24 in January 2024³
Underlying models	Google Gemini Pro and Gemini Nano, plus Google's Imagen 2 on Vertex AI for text-to-image generation⁴
Architecture	Hybrid: translations and Gallery Edit Suggestions run on-device; auto formatting, summarizing, corrections and image editing use the cloud⁵
Device reach	More than 200 million devices in 2024; about 400 million by 2025; Samsung targets 800 million in 2026⁶
Cost	Pricing after 2025 is not specified in the cited reporting; Samsung stated the features were free only until the end of 2025⁷
Privacy control	Users can disable cloud processing via Settings > Galaxy AI > "Process data only on device"⁵

Launch and initial features (January 2024)

Samsung unveiled the Galaxy S24 Ultra, S24+, and S24 in January 2024 as its first "AI phone", with features spanning call and text translation, the ProVisual Engine for imaging, and new search³^•². The launch set included:

Live Translate and Interpreter: real-time call and text translation in the Phone app covering 13 languages, one of the few Galaxy AI features that works without an internet connection⁸.
Circle to Search: searching anything on screen by drawing around it; it requires Android 14 with One UI 6.1 or higher and may not work in apps that block screen capture, such as video streaming or banking apps⁵.
Generative Edit: analyzes photo components and lets users move, remove, or transform elements, filling empty canvas areas; output is a resized photo of up to 12MP, requires an internet connection, and Samsung watermarks the results so altered images can be identified⁵^•⁸.
Note Assist: AI-generated summaries, template creation with pre-made formats, and cover creation in Samsung Notes³.
Transcript Assist: uses AI and Speech-to-Text to transcribe, summarize, and translate voice recordings with multiple speakers³.
Chat Assist: message composition and tone assistance, part of the S24 launch lineup².

How it works: hybrid architecture and underlying models

Galaxy AI splits work between the phone and the data center. According to Samsung's support documentation, Translation, time and weather Ambient wallpapers, and Google Messages features (Magic Compose, message suggestions, and rewrite suggestions) run on the device, while functions such as auto formatting, auto summarizing, auto corrections, and photo editing rely on the cloud⁵. Most translations are handled on the device itself, and Edit Suggestions within the Gallery are on-device, while most other features require a network connection⁵.

Why the split exists. Real-time call translation runs locally because latency matters too much for a cloud round-trip in live conversation⁹. A 2026 peer-reviewed survey identifies two broad obstacles to running large language models entirely on devices: prohibitive on-device resource requirements and the inherent limitations of cloud-centric approaches, which is the trade-off hybrid designs like Galaxy AI navigate¹⁰. Qualcomm's whitepaper on hybrid AI argues that the installed base of billions of AI-capable phones and PCs makes significant on-device generative processing already practical and poised to grow¹¹.

Models. At launch the suite was built on Google Gemini Pro and Gemini Nano, with Google's Imagen 2 on Vertex AI for text-to-image generation⁴. Reporting differs on the role of Samsung's own models: one account says Samsung uses Gemini Nano for local and offline features instead of its own Gauss models², while another describes Galaxy AI as a hybrid of Samsung's on-device models and Google's Gemini²; Reuters characterizes the features as "largely powered by Google's Gemini", alongside Samsung's own Bixby assistant⁶. For image tasks, Google uses Imagen 3 rather than Gemini, and Samsung uses the same model for its Photo Edit feature¹². By the Galaxy S26 launch in 2026, Gemini 3 series models were doing the heavy lifting, understanding goals, asking follow-up questions, and proactively suggesting next tasks¹³.

Privacy and user controls

Because some features send data to servers, Samsung gives users a switch: in Settings, under Galaxy AI, "Process data only on device" disables all cloud-based processing⁵. According to the S24 launch announcement, Galaxy S24 owners can decide exactly how much of their data is used to improve AI features, with Advanced Intelligence settings offering the option to turn off cloud-based data processing for those features.³ Samsung has also stated that data belonging to users and handled by Galaxy AI will neither be retained nor employed to train AI models unless the user expressly consents².

Samsung publishes per-feature processing-location details rather than blanket on-device claims⁹. The comparison with rivals is instructive. Apple's Private Cloud Compute sends only task-relevant data to Apple silicon servers, where data is not stored or made accessible to Apple, a claim the company has opened to independent security researcher verification¹²^•⁹. Gemini interactions, by contrast, are stored on Google's servers and used to train future language models, and opting out of training costs chat history and Extensions access¹². On the S26 generation, the Personal Data Engine learns from user preferences on device, Knox Enhanced Encrypted Protection (KEEP) isolates data within individual apps, and Knox Vault secures passwords, security keys, and biometrics in dedicated hardware¹⁴. One UI 9 adds a Security Brief to Now Brief with alerts for malware, apps from unknown sources, and unnecessary permissions, plus on-device Privacy Alerts for background permission-access attempts¹⁵.

Device support and rollout

Galaxy AI was first introduced with One UI 6.1, and features were available through One UI 8⁸. After launching on the Galaxy S24 series in January 2024, Samsung rolled the same features to the Galaxy S23 series just two months later, and compatibility now covers everything from the Galaxy S25 Ultra to the mid-range Galaxy A55¹⁶. Support extends to the Galaxy S23 series, Z Fold5/Flip5, and Tab S9 series via One UI 6.1, though the Tab S9 FE does not support AI features¹⁷. Supported hardware spans the S24/S25/S26 phone series, Tab S9 through Tab S11 tablets, and Galaxy Buds2 through Buds4 earbuds¹⁷.

Why older devices get a partial set. The S22 series and Z Fold4/Flip4 are supported with an incomplete feature set, and features like Audio Eraser and Natural Language Search do not work on the S23 series, 5th-generation foldables, or Tab S9 series because older NPUs cannot deliver the required on-device AI⁸. Which? likewise lists Galaxy AI as available on the Galaxy S23 series and newer S series devices and the Galaxy Z Flip/Fold 5 series and newer Z series devices⁷.

By the numbers

Samsung brought Galaxy AI to more than 200 million devices in 2024, beginning with the S24 series, under an "AI for All" strategy that included previous-generation devices¹⁸^•¹⁶.
In July 2025 Samsung committed to expand Galaxy AI to over 400 million devices globally by the end of 2025¹⁸. By January 2026, Reuters reported the company had rolled out Gemini-backed Galaxy AI features to about 400 million mobile products including smartphones and tablets, and plans to reach 800 million in 2026⁶. At CES 2026, co-CEO Roh Tae-moon framed the 800 million cumulative figure for 2026 under a "multi-AI engine" strategy¹.
Samsung reports that over 70% of Galaxy S25 users actively engage with AI-powered features developed with Google, and that Google Gemini usage tripled across the latest Galaxy S series¹⁸. A Samsung UK executive stated that over 5 million people in the UK use Galaxy AI regularly, a figure he compared to the internet taking eight years to reach¹⁶.
Independent context: Counterpoint Research reported in October 2025 that cumulative GenAI smartphone shipments surpassed 500 million units, and projected in June 2026 that GenAI smartphones would rise to 45% of global smartphone shipments in 2026¹⁹.

How it compares with Apple Intelligence and Pixel AI

The three suites differ in reach, privacy architecture, and default behavior. Apple Intelligence is locked to the iPhone 15 Pro series or newer, with the iPhone 16 series supporting it across the board¹², whereas Samsung and Google bring AI features to older or less expensive devices¹². Samsung leads by a wide margin in the number and maturity of individual AI features compared with Apple's Siri, per the Korea Herald's comparison¹.

Architecturally, Galaxy AI is the most explicitly hybrid of the three, combining Samsung's own on-device processing with a partnership on Google's Gemini models for cloud-side tasks⁹. Google's Pixel approach routes more open-ended requests to cloud Gemini by default rather than exhausting on-device options first⁹. In head-to-head testing, Tom's Guide's AI Phone Face-Off placed Galaxy AI first overall, with Apple third, crediting Galaxy AI's combination of Gemini's assistant powers, including cross-app actions, with Samsung-made apps for greater user freedom²⁰. Apple Intelligence's deep OS-level integration, however, makes it more useful day-to-day than Gemini in that outlet's assessment¹².

Recent developments (2024–2026)

2025. The Galaxy S25 series introduced Now Brief, deeper Gemini integration, and expanded multimodal capabilities, with some features extended to Galaxy S22 and S21 devices². The 2025 tools began rolling out to other phones from April 2025 onwards; Cross App Actions via Gemini work only on the S25 series, with native Google and Samsung apps plus major third-party apps like WhatsApp and Spotify⁸. Audio Eraser, new with the S25, identifies six audio types including voices, wind, and crowd noise⁴.

2026. The Galaxy S26 series, launched February 2026, is called by Samsung the "third-generation AI phone" and brings a multi-agent AI system powered by Perplexity and Google's Gemini, Now Nudge, and advanced Scam Detection, with Galaxy AI extended to Galaxy Buds4, Galaxy Watch, and Galaxy Tab devices². Now Nudge understands conversational context and surfaces relevant suggestions without app switching¹⁴. Google's Scam Detection, powered by an on-device Gemini model, is available in the Samsung Phone app on Galaxy S26 devices, giving instant audio and haptic alerts when a potential scam is detected during a call; analysis is entirely on-device and the feature is automatically off for contacts¹³. Samsung and Google also previewed an agentic Android platform powered by Gemini 3, launching on the S26 series as an early Google Labs preview¹⁴. Circle to Search, first introduced at Galaxy Unpacked 2024, now supports searching for multiple objects within a single image¹⁴. On September 16, 2026, Samsung began rolling out One UI 9 starting with the Galaxy S26, S26+, and S26 Ultra, following its debut on the Galaxy Z Fold8 Ultra, Fold8, and Flip8¹⁵. Samsung's stated aspiration is to move Galaxy AI beyond an eclectic bundle of features into an agentic "true AI companion" built on its Personal Data Engine⁸.

References

Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › AI companies, people, and products › AI products and assistants › AI in devices and work software

Initially written Sep 17, 2026 · Reviewed: Oct 15, 2026 · Edited: Oct 11, 2026; Oct 15, 2026 · Last review: Oct 15, 2026

Notice something wrong?