1. Home
  2. 🛠️ Developer Tools
  3. MLXHub
MLXHub: Local AI & LLM Server

MLXHub: Local AI & LLM Server

Private offline chat & tools

14 ratings
3 reviews
📊 Productivity
🛠️ Developer Tools
Free

Details

Released
August 19, 2026
Updated
October 21, 2026

Features

Developer

Juan Colilla
✨Promote your app ✨
MLXHub: Local AI & LLM Server screenshot #1 for iPhone
MLXHub: Local AI & LLM Server screenshot #2 for iPhone
MLXHub: Local AI & LLM Server screenshot #3 for iPhone
MLXHub: Local AI & LLM Server screenshot #4 for iPhone
MLXHub: Local AI & LLM Server screenshot #5 for iPhone
MLXHub: Local AI & LLM Server screenshot #6 for iPhone
MLXHub: Local AI & LLM Server screenshot #7 for iPhone
MLXHub: Local AI & LLM Server screenshot #8 for iPhone
MLXHub: Local AI & LLM Server screenshot #9 for iPhone
MLXHub: Local AI & LLM Server screenshot #10 for iPhone
iphone
ipad
⭐3
Reviews↘︎
📝9+ tags
Tags↘︎
🔄6+ apps
Alternatives↘︎
🎪2 available
Events↘︎
❓8 answered
FAQ↘︎
📦~ in 1 week
Updates frequency↘︎
🖼️Get Icon
Icons↘︎

About

Run powerful AI language models privately on your iPhone or iPad. Your data stays on your device, with options for distributed inference across multiple devices and a local LAN server for API access.

✓Run AI models locally on device
✓Distributed inference across devices (experimental)
✓Chat with LLMs and VLMs
✓Customizable AI personality and system instructions
✓Browse and install models from catalog or HuggingFace
✓Local LAN server for API access
✓Optimized for Apple Silicon
Show original description

What's New in MLXHub

2.3.0

October 21, 2026

You can talk to it now. Turn on voice mode, say what you need and hear the answer out loud. Transcription happens on your device, understands the languages you have set up, and you can cut in mid-sentence to take your turn. It also creates images without leaving your device. Download an image model from the new Image tab, or ask the assistant in the middle of a conversation: it suggests the image and you adjust the description and size before creating it. You can pause it and pick it up later, and every image stays in your gallery. A new benchmark measures the real speed of your iPhone or iPad with real models. Publish your score to the leaderboard and see where it lands against other devices. The assistant has a new face. The Entity listens, thinks and answers with you in chat, in voice mode and on the server, and it brings a new icon along. On iPad the app makes the most of the screen with a sidebar, and on iPhone Duo it adapts to book and tabletop postures. The server on your local network talks too: open its voice center and ask the model it serves out loud, with the screen dimmed while it waits. The web chat portal has been redesigned and keeps its history in each browser. By default, the server only answers your local network. If you are starting fresh, the new first-run flow asks whether you came to chat or to use the developer tools, and recommends a first model for your device. Settings and Network have been reorganized so you find things the first time, model search shows real sizes and more results, and downloads can be followed from the Lock Screen. Compatibility depends on the model you choose and the device you have. Every model that fits your device, images up to 1280 px, write tools, Skills and hosting a mesh are part of MLXHub Plus. The LAN server and its voice center unlock with Plus or with a one-time purchase. Thanks for being here.

More

Events

Read more about Talk to It Under the Stars
AvailableMajor Update

Talk to It Under the Stars

Voice mode is here: just say what you need

Nov 1, 2026 — Nov 15, 2026

Read More

Read more about Nodes, Assemble!
Event EndedMajor Update

Nodes, Assemble!

Distributed inference across your devices

Aug 27, 2026 — Sep 19, 2026

Read More

See All

User reviews

Most RecentHighest RatedLowest Rated

还行

就是AI下载太慢了
小小狗78

Tragically Underrated

Best way to host local models from an IPad.
-B-CHiNO BLaST-

Response from developer

Thank you so much for those words! Your review is the fuel for this app development! Hope you can get the latest features ASAP (as Apple is holding the app down from being released)… Keep an eye on it!

Does what it says.

Needed a quick way to benchmark smaller Gemma models on device and this did the job without requiring me to wire up a test harness. The pricing disconnect is steep. Charging a $40/year subscription (or a $240 'lifetime' pledge) for what essentially feels like a lightweight wrapper around existing open-source frameworks and inference runtimes is tough to justify. The developer clearly cares, but this is a utility tool being priced like an enterprise SaaS platform.
Show more
ZZimzooma

Response from developer

Hello Zimzooma! Thank you for your comment, I really appreciate every opinion over the app, even the ones that don’t think it is perfect (because they allow me to improve it!). I totally agree with your opinion, let me clarify myself: The price is not for the tool as it is today, but for the heavy workload it has under the hood. For MLXHub to exist a lot of work has to be done, not only in the app, but in the forks of Open Source frameworks from Apple that don’t allow most of the new features MLXHub has (and that soon will be in production). Next version of MLXHub will allow you to run Qwen3.6 35B A3B in a single iOS device, as well as Bonsai 2 27B on supported ones. With tons of improvements coming to make MLXHub the BEST private AI chat for iOS and even compatibility with embedding models and more standards. The price of the subscription is justified with the continuous development that you get for no extra money, and the lifetime pays itself for those users who heavily rely on the app long-term! I suggest you to follow @MLXHub on X to be aware of all the new things, and contact support if you have any issue or request. As a solo-developer, it is my pleasure to help my users take full benefit of the app!

Developer apps

agendy: Contact Cleaner
agendy: Contact Cleaner

Swipe Clean. Private. One Pay.

📊 Productivity

(1)

Free
In-App Purchases
Poolendar
Poolendar

IBS & IBD Bowel Tracker

💪 Health & Fitness
Free
In-App Purchases
See developer apps

FAQ

What is MLXHub?

MLXHub allows you to run open-source language models and vision-language models directly on your iPhone and iPad. It prioritizes privacy by keeping all model inference and conversations on your device.

Can I run MLXHub models offline?

Yes, MLXHub is designed for offline use. Model inference happens entirely on your device, meaning you do not need an internet connection to chat with downloaded models.

What is distributed inference in MLXHub?

Distributed inference is an experimental feature that lets you pool memory from multiple Apple Silicon devices on the same Wi-Fi network to run larger models than any single device could handle. Hosting this feature requires MLXHub Plus.

How does MLXHub handle privacy?

MLXHub ensures privacy by performing all model inference on your device, meaning your data never leaves. It does not require an account and is ad-free. Anonymous diagnostic data is collected but can be turned off.

Is MLXHub free to use?

MLXHub is free to download and use for running AI models locally and chatting. Features like hosting a distributed inference mesh and the local LAN server are part of MLXHub Plus.

What devices are supported by MLXHub?

MLXHub is available on iPhone and iPad devices. It is built to leverage Apple Silicon for optimal performance, with specific features like Apple Intelligence integration on iPhone 15 Pro/16 and later.

How often is MLXHub updated?

The latest version of MLXHub is 2.0.0, last updated on August 25, 2026. Updates typically focus on improving model compatibility, performance, and adding new features to the local AI experience.

Can I use MLXHub models with other applications?

Yes, MLXHub can function as a local LAN server (part of MLXHub Plus) that exposes an API compatible with /v1/chat/completions. This allows other applications on your local network, such as a Mac running OpenCode or custom scripts, to utilize your device's models.