Title: Google AI Edge Gallery | BestBlogs.dev
URL Source: https://www.bestblogs.dev/article/87b99a3f
Published Time: 2026-04-06 05:18:26
Markdown Content: Skip to main content Toggle navigation menu Toggle navigation menuArticlesPodcastsVideosTweetsSourcesNewsletters
⌘K
Change language Switch ThemeSign In
Narrow Mode
Google AI Edge Gallery
S Simon Willison's Weblog @Simon Willison
One Sentence Summary
Google's official AI Edge Gallery app brings local Gemma 4 and Gemma 3 models to the iPhone, featuring fast performance, multimodal capabilities, and interactive tool-calling skills.
Summary
Simon Willison reviews Google's AI Edge Gallery, an official iOS app designed to run Gemma 4 (E2B and E4B sizes) and Gemma 3 models locally on mobile hardware. The app demonstrates impressive performance, with the 2.54GB E2B model providing fast and practical utility. Key features include image-based Q&A, short audio transcription, and a unique 'skills' demo that showcases tool calling against eight interactive HTML widgets, such as maps and hash calculators. While the app marks a significant milestone as the first official local model app from a major vendor, it currently lacks persistent conversation logs and exhibits some stability issues in complex demos.
Main Points
* 1. Google has launched an official iOS app for local execution of Gemma models.This represents the first time a major model vendor has released a dedicated mobile app for running their LLMs locally on consumer hardware, specifically targeting the Gemma 4 and Gemma 3 families. * 2. The Gemma 4 E2B model offers a balance of size and performance on mobile.At 2.54GB, the model is small enough for mobile storage while remaining fast and useful for on-device tasks without requiring cloud connectivity. * 3. The app features a sophisticated 'skills' demo for tool calling.It demonstrates the model's ability to interact with HTML-based widgets (e.g., interactive maps, QR code generators), showcasing the potential for local agentic workflows. * 4. Current UX limitations include ephemeral conversations.The app does not yet support permanent logs or conversation history, and the tool-calling demo can occasionally freeze during follow-up prompts.
Metadata
AI Score
86
Website simonwillison.net
Published At Today
Length 171 words (about 1 min)
Sign in to use highlight and note-taking features for a better reading experience. Sign in now
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.
(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.
S Simon Willison's Weblog @Simon Willison
One Sentence Summary
Google's official AI Edge Gallery app brings local Gemma 4 and Gemma 3 models to the iPhone, featuring fast performance, multimodal capabilities, and interactive tool-calling skills.
Summary
Simon Willison reviews Google's AI Edge Gallery, an official iOS app designed to run Gemma 4 (E2B and E4B sizes) and Gemma 3 models locally on mobile hardware. The app demonstrates impressive performance, with the 2.54GB E2B model providing fast and practical utility. Key features include image-based Q&A, short audio transcription, and a unique 'skills' demo that showcases tool calling against eight interactive HTML widgets, such as maps and hash calculators. While the app marks a significant milestone as the first official local model app from a major vendor, it currently lacks persistent conversation logs and exhibits some stability issues in complex demos.
Main Points
* 1. Google has launched an official iOS app for local execution of Gemma models.
This represents the first time a major model vendor has released a dedicated mobile app for running their LLMs locally on consumer hardware, specifically targeting the Gemma 4 and Gemma 3 families.
* 2. The Gemma 4 E2B model offers a balance of size and performance on mobile.
At 2.54GB, the model is small enough for mobile storage while remaining fast and useful for on-device tasks without requiring cloud connectivity.
* 3. The app features a sophisticated 'skills' demo for tool calling.
It demonstrates the model's ability to interact with HTML-based widgets (e.g., interactive maps, QR code generators), showcasing the potential for local agentic workflows.
* 4. Current UX limitations include ephemeral conversations.
The app does not yet support permanent logs or conversation history, and the tool-calling demo can occasionally freeze during follow-up prompts.
Key Quotes
* This is Google's official app for running their Gemma 4 models... directly on your iPhone. * The E2B model is a 2.54GB download and is both fast and genuinely useful. * This is the first time I've seen a local model vendor release an official app for trying out their models on in iPhone. * Conversations with this app are ephemeral.
AI Score
86
Website simonwillison.net
Published At Today
Length 171 words (about 1 min)
Tags
Google Gemma
Local LLM
iOS
Mobile AI
Edge Computing
Related Articles
* Introducing Moltworker: a self-hosted personal AI agent， minus the minis * My fireside chat about agentic engineering at the Pragmatic Summit * Clawdbot/moltbot Clearly Explained (and how to use it), an AI agent framework that acts as an autonomous digital employee for solopreneurs to automate coding, research, and business operations.") * First impressions of Claude Cowork， Anthropic’s general agent * Introducing Showboat and Rodney， so agents can demo what they’ve built * GPT-5.4 mini and GPT-5.4 nano: 76,000 Photos for $52 * Bring state-of-the-art agentic skills to the edge with Gemma 4 * Slashing agent token costs by 98% with RFC 9457-compliant error responses * GPT-5.4 Makes A Splash， AI’s Growth on Mobile， Data Centers Go Off-Grid， and more... * Wilson Lin on FastRender: a browser built by thousands of parallel agents HomeArticlesPodcastsVideosTweets