Pocket LLM Chat

Pocket LLM Chat

Pocket LLM Chat Tool offline

0 ratings
Free

Rating summary

Details

Released
September 1, 2026
Updated
November 15, 2026

Features

Pocket LLM Chat screenshot #1 for iPhone
Pocket LLM Chat screenshot #2 for iPhone
Pocket LLM Chat screenshot #3 for iPhone
Pocket LLM Chat screenshot #4 for iPhone
🖼️Get Icon
Icons↘︎

About

Pocket LLM is a private, on-device AI chat app for iPhone. Everything runs locally on your device using Apple MLX—no accounts, no cloud, no tracking. Download a compact model once, then chat offline whenever you want. Your prompts, conversations, images, and replies never leave your iPhone. TWO COMPACT MODELS LFM3 1.7B (4-bit) — fast, capable text chat Bonsai 2 2B (2-bit) — compact chat with image understanding FAST, NATURAL CHAT Streaming replies appear as they are written Receive prompts from other apps, Shortcuts, or scripts via pocketllm:// links Full Markdown rendering for headings, lists, code blocks, and links Send photos with or without text to a vision-capable model Optional tool calling with a built-in calculator Conversation history is saved automatically per model Export any conversation as Markdown Clear chats and manage downloaded models and storage from the app PRIVATE BY DESIGN All inference happens on your device No sign-in, analytics, telemetry, crash reporting, ads, or cross-app tracking Works fully offline after models are downloaded Only model downloads use the network; choose Hugging Face or a mirror in the app GOOD TO KNOW The first launch downloads a model (about 1–1.2 GB); Wi-Fi is recommended Performance and memory use depend on your iPhone Model responses can be inaccurate; verify important information Some features, such as image input, require a compatible model Requires iOS 18 or later Pocket LLM is designed for private everyday questions, writing help, brainstorming, and learning—without sending your data to a server.
Show more
+2

What's New in Pocket LLM Chat

1.2.1

November 15, 2026

If the model comes back with nothing, Pocket LLM retries with a shorter instruction instead of failing immediately, so a one-off request is far more likely to be answered. When no answer comes back at all, the message names the model that was tried and how many attempts were made, which makes a problem easy to report. Unchanged: requests run entirely on-device while Pocket LLM is in the foreground, and your own chat history is never touched.

More

Developer apps