content image

Google の Gemma 4 は、あなたのスマートフォンを AI の強力マシンにしようとしている——クラウド不要で

Google's Gemma 4 Wants to Make Your Phone an AI Powerhouse—No Cloud Required

GoogleがGemma 4を発表し、クラウド不要でスマホ上で高速・省電力に動くオンデバイスAIの実現に本格的に乗り出した。対応機種の拡大と開発者ツールの整備で、次の主戦場となるか。
分からないところをタップすると
↓日本語訳が表示されます↓

For years, “AI on your phone” often meant that your phone was only a window to a cloud model. Google’s April 2, 2026 rollout of Gemma 4 suggests a more ambitious idea: AI that can run locally, quickly, and privately on the device itself. In Android’s AICore Developer Preview, Google offers two mobile-focused Gemma 4 models, E2B and E4B, and says they are the foundation for the next generation of Gemini Nano 4. Google also says Gemma 4 supports more than 140 languages, understands text, images, and audio, and can be up to four times faster while using up to 60% less battery than previous versions. (android-developers.googleblog.com)

What makes this important is not only the model, but the system around it. Gemini Nano runs through Android’s AICore service, which manages model updates, uses on-device hardware for lower latency, and keeps requests local instead of sending them to servers. Google’s ML Kit GenAI APIs are built on top of this layer, so developers can add features like summarization, rewriting, image description, speech recognition, and custom prompting without building everything from zero. Just as importantly, support is no longer limited to Pixel phones: Google’s latest documentation lists devices from Samsung, Xiaomi, OPPO, vivo, Honor, Motorola, and others. (developer.android.com)

Google is also trying to prove that “local AI” can do more than answer questions. In Google AI Edge Gallery, available on Android and iOS, Gemma 4 powers “Agent Skills” that can summarize content, create flashcards and graphs, and connect with tools such as Wikipedia. A separate Google demo shows fully offline function-calling on mobile, such as creating calendar events or opening maps, which brings the idea of a real on-device assistant much closer. (developers.googleblog.com)

Can Google turn on-device AI into its next main battlefield? The signs are strong, because it now has the model, the OS layer, the developer tools, and a widening hardware ecosystem. Still, Google admits that the current preview works best on AICore-enabled devices, while unsupported phones may fall back to CPU performance that is “not representative” of the final experience. In other words, Google has momentum, but the real test will be whether this smooth, offline AI experience reaches ordinary smartphones at scale. (android-developers.googleblog.com)

by EigoBoxAI
作成:2026/04/15 15:01
レベル:中上級 (語彙目安:4000〜6000語)

まだ読んでいないコンテンツ

content image
by EigoBoxAI
作成:2026/04/19 15:12
レベル:中上級 (語彙目安:4000〜6000語)
content image
by EigoBoxAI
作成:2026/04/19 09:04
レベル:初級 (語彙目安:300〜1000語)
content image
by EigoBoxAI
作成:2026/04/19 09:03
レベル:中級 (語彙目安:2000〜2500語)
content image
by EigoBoxAI
作成:2026/04/19 09:01
レベル:初中級 (語彙目安:1000〜2000語)
content image
by EigoBoxAI
作成:2026/04/19 03:09
レベル:超上級 (語彙目安:8000語以上)
content image
by EigoBoxAI
作成:2026/04/19 03:07
レベル:上級 (語彙目安:6000〜8000語)
content image
by EigoBoxAI
作成:2026/04/19 03:04
レベル:中上級 (語彙目安:4000〜6000語)
content image
by EigoBoxAI
作成:2026/04/18 21:04
レベル:超入門 (語彙目安:〜300語)
content image
by EigoBoxAI
作成:2026/04/18 21:03
レベル:初級 (語彙目安:300〜1000語)
content image
by EigoBoxAI
作成:2026/04/18 21:02
レベル:中級 (語彙目安:2000〜2500語)