Incept5

Incept5 / mlxchat

Public

A ChatGPT-like app using Alibaba's Qwen3.5 models locally on the phone - with tool calling & vision

49
9
100% credibility
Found Mar 07, 2026 at 49 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Swift
AI Summary

A native iPhone and iPad app for chatting with smart local AI models that handle text, images, and optional web lookups.

How It Works

1
🔍 Discover MLXChat

You hear about this handy chat app that lets you talk to a smart helper right on your iPhone or iPad, all without needing the internet.

2
📱 Launch the app

Tap to open the app and see the clean chat screen with tabs for conversation and settings.

3
⚙️ Tweak your setup

Head to settings to pick memory limits, turn on features like streaming replies, and add a search key for fresh web info if you like.

4
🧠 Choose your AI buddy

Browse the list of ready-to-go brains, pick one like the recommended size, and watch the colorful progress ring download and prepare it just for you.

5
💬 Start your first chat

Type a question or hello, hit send, and see words appear live as your new companion thinks and replies.

6
🖼️ Add photos or web smarts

Snap or pick a picture to analyze, share a web link to summarize, or ask for today's news – it handles images and fetches updates smoothly.

🎉 Enjoy endless talks

Now chat privately anytime with your offline AI friend, getting quick answers, ideas, or help tailored just how you set it up.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 49 to 49 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is mlxchat?

MLXChat is a ChatGPT-like app built in Swift that runs Alibaba's Qwen3.5 models locally on iPhone and iPad using Apple's MLX framework. It delivers a familiar chat interface for text conversations, image analysis, and tool calling—like web search via Brave API or fetching URL content—all offline and free. Developers get a ready-to-build iOS project with model downloading, streaming responses, and customizable system prompts including date, location, and device vars.

Why is it gaining traction?

Unlike cloud-dependent ChatGPT-like apps, this runs quantized Qwen3.5 models (0.8B to 9B) entirely on-device for privacy and zero latency, with smart prefetching for URLs or current events to boost tool accuracy. The clean, system-native UI ages well across iOS updates, and optional reasoning mode plus Markdown rendering make responses feel polished. Chinese models like Alibaba's Qwen shine here, rivaling bigger Western LLMs in a free, local package.

Who should use this?

iOS developers prototyping offline AI companions or studying on-device inference with vision and tools. Hardware enthusiasts on recent iPhones/iPads testing MLX limits with real apps. Teams forking for custom chatgpt-like apps for study aids or internal copilots, skipping API costs.

Verdict

Grab it if you're into local AI experiments—solid docs and a macOS test harness make iteration fast, despite 49 stars and 1.0% credibility signaling early maturity. Needs more models and perf tweaks for production, but it's a practical starter for phone-based LLMs.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.