Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Mobile & Edge LLMs

fullmoon Review: Free Offline AI Chat for iPhone, iPad and Mac

·5 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Warning: fullmoon has had no release since 28 January 2025 (version 1.2.0 on the App Store) and no code change since 5 May 2025, so it may be unmaintained — check that it works on your device before relying on it. It is a free chat app by Mainframe Computer, Inc. that runs small language models on an iPhone, iPad or Mac. Every statement here comes from the maker's website, the App Store listing, the privacy policy and the public repository.

Key Takeaways

  • Maintenance: the repository's last change added Qwen3 models, which the 1.2.0 release notes and the website do not list.
  • License: MIT per the LICENSE file; the website itself says only "open source".
  • Cost: free to download.
  • Maker: Mainframe Computer, Inc.; the App Store listing is English only.
  • Devices: iOS 17.6 or later; the website also names visionOS.

📍 In One Sentence

fullmoon is a free, MIT-licensed iPhone, iPad and Mac app that chats with small language models running on the device through Apple's MLX.

💬 In Plain Terms

You pick one of a few small models and chat with it like any assistant, with the work done on your own device instead of a cloud service.

📌Note: This review is based on the website, the App Store listing, the privacy policy and the public repository, checked on 30 October 2026. PromptQuorum has not installed or benchmarked the app, so answer quality and speed are not assessed.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

fullmoon is a free chat app from Mainframe Computer, Inc. that runs small language models on an iPhone, iPad or Mac through Apple's MLX framework and says it works fully offline. Its source code is published under the MIT license on GitHub. This review is based on the maker's website, the App Store listing, the privacy policy and the repository, checked on 30 October 2026; PromptQuorum has not tested the app hands-on.

fullmoon Review: Free Offline AI Chat for iPhone, iPad and Mac

What Is fullmoon?

fullmoon is a chat app for talking to language models that run on your own Apple device instead of on a server. The App Store lists it as "fullmoon: local intelligence" in Utilities from the seller Mainframe Computer, Inc., first released on 13 January 2025, and the repository describes it as an app to "chat with private and local large language models".

The repository credits Apple's MLX Swift for the inference, and the website says the app is optimized for Apple silicon. The code repository was created in October 2024.

Get It

fullmoon is a free App Store download for iPhone, iPad and Mac; the website names no Android or Windows version.

Option
How to get it
iPhone, iPad, Macfullmoon on the App Store, free
Newer buildsTestFlight, for newer builds
Websitefullmoon.app
Privacy policyfullmoon.app/privacy
Source codeGitHub, MIT license

Version reviewed: 1.2.0, released on 28 January 2025 per the App Store; the app is 31.6 MB before any model download, needs iOS 17.6 or later, is rated 17+ and is listed in English only. GitHub shows releases up to 1.2.3 dated 27 January 2025, a day before the App Store date, so the two version numbers do not line up and this review uses the App Store version.

Features Confirmed by the Sources

Every item below comes from the website, the App Store description or the repository; none has been independently tested.

  • On-device chat. The models run on the device and the app works without an internet connection, per the website; chat history is stored locally, per the README.
  • Model list. The website names Llama 3.2 1B (0.7 GB) and 3B (1.8 GB) and DeepSeek-R1-Distill-Qwen-1.5B in 4-bit (1.0 GB) and 8-bit (1.9 GB). The repository's main branch also lists Qwen3 4B and 8B in 4-bit; whether the 1.2.0 App Store build includes them is not stated.
  • Where models come from. The sources do not say how the models are downloaded; the model IDs in the code point to the mlx-community collection on Hugging Face.
  • Personalization. The theme, fonts and system prompt can be changed in the app.
  • Shortcuts. An action lets you chat with a local model or hand its output to other Shortcuts actions.

Data Flow and Privacy

The privacy policy, effective 23 December 2024, says "We do not collect any information from you" and "We do not share your personal information with third parties". The website calls the app private and says it runs models locally; the README says chat history is saved locally.

The same policy is not consistent, though: it also lists what "the information we collect" is used for, including to "understand and analyze how you use fullmoon", and ends with consent to "our collection, use, and sharing of your information". It does not say what, if anything, is collected, and it does not mention on-device processing at all.

📌Note: These are the maker's declarations, not audit results. The source code is public, but PromptQuorum has not read it or inspected the app's network traffic.

Who Should Use It

  • Apple-device owners who want the simplest possible offline chat and are happy with small models.
  • Developers and tinkerers who want an MIT-licensed Swift app built on MLX Swift to study or fork.
  • Anyone comparing offline chat apps for iPhone who values open source over frequent updates.

What We Could Not Verify

  • Maintenance status. The warning at the top of the page applies; PromptQuorum found no statement from the maker on whether development continues.
  • Answer quality and speed. Not tested; the sources give model sizes but no RAM needs or tokens per second, so which iPhones run which model is unknown.
  • Model licences. The models carry their own licences (the website links the Meta Llama and DeepSeek model pages); PromptQuorum has not reviewed them.

Competitors and Alternatives

Private Mind

Platforms:
iOS, Android
Price / license:
Free / MIT
Key difference:
Adds on-device document Q&A and benchmarks; runs on ExecuTorch rather than MLX
Articles about Private Mind (2)

Also mentioned in:

+9 more not shown

fullmoon is the only one of the four without an Android version; confirm each app's current price, license and platforms on its own listing.

Compare in directory

Frequently Asked Questions

Can I try newer builds than the App Store version?

The website offers a TestFlight link for the latest features and models, which may include models the App Store build lacks. PromptQuorum did not join it.

Does fullmoon work on a Mac?

The website lists macOS and the App Store listing shows Mac compatibility. PromptQuorum did not check whether it is a native Mac app or the iPad app running on Apple silicon.

Does it cost anything outside the United States?

On the App Store storefronts checked on 30 October 2026 (Germany, France, Spain, Japan, South Korea and Brazil) it was free and showed the same version, 1.2.0.

Can I load my own models?

The sources describe a fixed in-app model list and do not mention importing other models.

Verdict

fullmoon is a clean, free and open-source way to chat with small models offline on an iPhone, iPad or Mac, with useful touches such as a Shortcuts action and an editable system prompt. Against that, it has had no release since January 2025, its privacy policy contradicts itself, the model list is short and the quality and speed of its answers could not be checked here. It suits people who value a simple, MIT-licensed app and accept that it may no longer be maintained; anyone who needs more than basic chat, or regular updates, should compare the alternatives above first.

Sources

← Back to Power Local LLM

MCP