AI Navigate
インサイト最新記事一覧AI大全

Google AI Edge Gallery

Simon Willison's Blog / 4/6/2026

📰 NewsSignals & Early TrendsTools & Practical UsageModels & Research
Read original →
共有:

Key Points

  • Google has released an iPhone app, “Google AI Edge Gallery,” to run its Gemma 4 models locally on-device, including E2B and E4B sizes (plus some Gemma 3 family models).
  • The app is reported to work very well in practice, with the Gemma 4 E2B model downloading at about 2.54GB and delivering fast, genuinely useful results.
  • It supports multimodal features such as “ask questions about images” and short audio transcription (up to 30 seconds) using the smaller Gemma 4 models.
  • A showcased “skills”/tool-calling demo lets the model call interactive widgets implemented as HTML pages (e.g., interactive map, hash calculation, Wikipedia query, QR code), though the source code is not provided.
  • The reviewer notes some limitations, including missing permanent conversation logs and an issue where a follow-up prompt can freeze the app.
Sponsored by: WorkOS — Production-ready APIs for auth and access control, so you can ship faster.

6th April 2026 - Link Blog

Google AI Edge Gallery (via) Terrible name, really great app: this is Google's official app for running their Gemma 4 models (the E2B and E4B sizes, plus some members of the Gemma 3 family) directly on your iPhone.

It works really well. The E2B model is a 2.54GB download and is both fast and genuinely useful.

The app also provides "ask questions about images" and audio transcription (up to 30s) with the two small Gemma 4 models, and has an interesting "skills" demo which demonstrates tool calling against eight different interactive widgets, each implemented as an HTML page (though sadly the source code is not visible): interactive-map, kitchen-adventure, calculate-hash, text-spinner, mood-tracker, mnemonic-password, query-wikipedia, and qr-code.

Screenshot of an "Agent Skills" chat interface using the Gemma-4-E2B-it model. The user prompt reads "Show me the Castro Theatre on a map." The model response, labeled "Model on GPU," shows it "Called JS skill 'interactive-map/index.html'" and displays an embedded Google Map centered on a red pin at The Castro Theatre in San Francisco, with nearby landmarks visible including Starbelly, Cliff's Variety, Blind Butcher, GLBT Historical Society Museum, and Fable. An "Open in Maps" link and "View in full screen" button are shown. Below the map, the model states "The interactive map view for the Castro Theatre has been shown." with a response time of 2.4 s. A text input field with "Type prompt..." placeholder, a "+" button, and a "Skills" button appear at the bottom.

(That demo did freeze the app when I tried to add a follow-up prompt though.)

This is the first time I've seen a local model vendor release an official app for trying out their models on in iPhone. Sadly it's missing permanent logs - conversations with this app are ephemeral.

Posted 6th April 2026 at 5:18 am

This is a link post by Simon Willison, posted on 6th April 2026.

google 401 iphone 63 ai 1950 generative-ai 1731 local-llms 153 llms 1698 gemini 182 llm-tool-use 66

Monthly briefing

Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.

Pay me to send you less!

Sponsor & subscribe

💡 Insights using this article

This article is featured in our daily AI news digest — key takeaways and action items at a glance.

📅 4/6DailyView insight →

関連おすすめサービス

※当サイトはアフィリエイト広告を利用しています
🎧

Notta搭載AI議事録イヤホン ZENCHORD1

AI時代の仕事術。Notta搭載で会議の議事録を自動生成するスマートイヤホン。

🎙️

AI搭載ボイスレコーダー Plaud

世界100万人が愛用。AIで文字起こし・要約を自動化するボイスレコーダー。

🖼️

画像高画質化AIツール Aiarty Image Enhancer

AIで画像を高画質化。写真・イラストを簡単にアップスケール。

Google AI Edge Gallery | AI Navigate