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

Ensu Review: Free On-Device Chat App from Ente Labs

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

Ensu (version 0.1.22, released on 23 October 2026) is a free chat app from Ente Labs that runs downloaded language models on the device, with no account. It is open source, and the maker itself says it is not yet as capable as ChatGPT or Claude Code. Every statement here comes from public sources, not from our own tests.

Key Takeaways

  • Maintenance: the releases page shows regular releases from April to October 2026, the latest on 23 October 2026, so no stale-project warning applies.
  • Models: downloaded inside the app; the source code lists seven, among them LFM 2.5, Qwen 3.5 and Gemma 4.
  • Cost: free with no account; the maker has not announced any pricing.
  • Platforms: iOS, Android, macOS, Windows and Linux, plus an experimental web version.
  • Openness: source code on GitHub in a repository licensed AGPL-3.0.

📍 In One Sentence

Ensu is an open-source chat app from Ente Labs that downloads a small language model and answers on your phone or computer instead of in the cloud.

💬 In Plain Terms

You install the app, let it download a model once, and then chat, attach images or dictate by voice without an account.

📌Note: This review is based on public sources (the maker's blog and site, the store listings, GitHub release notes and the source code), checked on 24 October 2026. PromptQuorum has not installed or benchmarked Ensu, so answer quality and speed are not assessed.

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

Ensu is a free, open-source chat app from Ente Labs that runs language models on your own phone or computer. It is available for iOS, Android, macOS, Windows and Linux, plus an experimental web version, and its code is on GitHub. This review is based on the maker's blog post and website, the store listings, the GitHub release notes and the source code, checked on 24 October 2026; PromptQuorum has not tested the app hands-on.

Ensu Review: Free On-Device Chat App from Ente Labs

What Is Ensu?

Ensu is a chat app from Ente Labs that the maker describes as "a chatgpt-like app that runs completely on your device with full privacy and zero cost". The announcement is dated 2 March 2026, the App Store lists the seller as Ente Technologies, Inc., and the first App Store release was on 27 February 2026.

According to the announcement, the core logic is written in Rust, the mobile apps are native and the desktop apps use Tauri; the source code runs the models with llama.cpp. The App Store title is "Ensu - Ente's Local LLM".

Get It

Ensu has an installer or store listing on every major system: App Store, Google Play and GitHub Releases downloads for macOS, Windows and Linux.

Option
How to get it
iPhone / iPadEnsu on the App Store, free
AndroidGoogle Play or the APK on GitHub Releases
macOSUniversal .dmg on GitHub Releases
Windowsx64 setup .exe on GitHub Releases
LinuxAppImage, .deb (amd64) or .rpm (x86_64) on GitHub Releases
Websiteente.com/ensu
Privacy policyente.com/privacy
Source codegithub.com/ente-io/ente, AGPL-3.0

Version reviewed: 0.1.22, released on 23 October 2026 on GitHub and the App Store; the iOS app is 61.6 MB before any model download. The maker labels the web version experimental. The Google Play page showed an update date of 7 October 2026, earlier than the 0.1.22 release; the APK on GitHub carries 0.1.22.

Models and Requirements

Ente does not publish a RAM requirement or a model list on its website; the models and default choices below come from the open-source code and the release notes.

Models

What the sources say:
Downloaded inside the app. The code lists seven: LFM 2.5 350M, LFM 2.5 VL 1.6B, Qwen 3.5 0.8B, 2B and 4B, and Gemma 4 E2B and E4B

Download size

What the sources say:
Main file from about 0.2 GB (LFM 2.5 350M) to about 5.0 GB (Gemma 4 E4B); vision models add an image file of about 0.2 to 1.0 GB

Default on phones

What the sources say:
By device memory: LFM 2.5 350M up to 4 GB, LFM 2.5 VL 1.6B above 4 GB up to 8 GB, Gemma 4 E2B above 8 GB; chat is disabled below 4 GB

Default on desktop

What the sources say:
Gemma 4 E4B with 16 GB of memory or more, otherwise LFM 2.5 VL 1.6B; the app offers a few other models to switch to

Operating systems

What the sources say:
iOS 16.0 or later per the App Store; Windows x64 and Linux amd64 / x86_64 packages; no minimum is stated for Android or macOS

Internet

What the sources say:
Needed to download models, packs and desktop updates; chatting afterwards works offline per the maker

The model policy was read from the main branch of the open-source code on 24 October 2026 and can change with any release; the phone limits come from the code and the 0.1.17 release notes.

Features Confirmed by the Sources

Every item below comes from the maker's announcement, the release notes or the source code; none has been independently tested.

  • Local chat. Conversations run on the device; the maker lists no web search and says the app is "not as powerful as ChatGPT or Claude Code".
  • Image attachments. Images can be attached to chats; phones resize them to the model's limits (0.1.16) and desktop accepts dragged-in images (0.1.17).
  • Voice dictation. On iOS and Android, speech for the chat box is transcribed locally with a separate downloadable speech model (0.1.16).
  • Ensu Packs. Optional knowledge packs for more accurate answers, added on mobile in 0.1.18 and on desktop in 0.1.20.
  • Your Notes on desktop. Chat with local markdown notes (0.1.20), with checks for changed or unavailable sources in follow-up questions (0.1.21).

Data Flow and Privacy

The maker says processing happens on the device and that Ensu offers "full privacy"; the App Store listing adds that it needs no account and has no tracking or usage limits. The announcement says end-to-end encrypted sync and backup through an Ente account, with self-hosting, are planned but not enabled in this release.

The code fetches models from Hugging Face repositories published by Ente and the speech model from models.ente.com, and desktop apps check for updates. Ente's privacy policy, last updated on 16 October 2026, does not mention Ensu; it says Ente collects no analytics about how app features are used and lists anonymized crash reports and Hugging Face for model hosting.

📌Note: These are the maker's statements and what the public code shows, not audit results. PromptQuorum has not inspected network traffic or stored data; because the code is public, others can check the claims.

Who Should Use It

  • Privacy-minded users who want a chat app without an account whose code anyone can read.
  • People with several devices who want the same app on iPhone, Android, Mac, Windows and Linux.
  • Tinkerers on modest hardware who accept small models for simple questions and chats over their own notes.

What We Could Not Verify

  • Answer quality and speed. PromptQuorum did not run the models, so quality, speed and battery use on any device are not assessed.
  • Model terms. The licences of the downloaded models and of the converted files Ente hosts were not reviewed; this is not legal advice.
  • Sync and backup. Not enabled in this release, so chats stay on the device where they were made until the planned sync ships.
  • Hardware needs. No RAM requirement is published; the memory thresholds above come from the source code.
  • Stability. The version is 0.1.22, the maker calls this a first release without pricing or stability commitments, and the web version is labelled experimental.

Competitors and Alternatives

Locally AI

Platforms:
iOS, macOS
Price / license:
Free / closed source
Key difference:
Runs open models through MLX on Apple devices; no Android or Windows
Articles about Locally AI (5)

Also mentioned in:

Compared from each app's PromptQuorum directory entry; confirm current platforms, price and licence on each app's own listing.

Compare in directory

Frequently Asked Questions

Is Ensu free, and will it stay free?

Ensu is free to use and the announcement calls it "zero cost", but the maker says it has not set pricing yet, so the sources cannot rule out later changes.

Can I bring my own models?

The sources do not say. The app downloads the models in its catalog, and none of the sources checked mention importing your own model files.

Can I continue a chat on another device?

Not yet. Sync is not enabled in this release; the maker plans optional end-to-end encrypted sync and backup through an Ente account and says existing local chats will be included once it arrives.

Why does the iOS link on the Ensu page say "ente-locker"?

The link on ente.com/ensu points to App Store id 6758197006, whose URL slug reads "ente-locker" while the listing itself is titled "Ensu - Ente's Local LLM". This review links the listing by its id.

Verdict

Ensu is a free, open-source way to run small language models on a phone or a computer without an account, from a maker that ships regular releases. Against that, it is a 0.1.x release with small models, sync and backup are not available yet, and none of its privacy, speed or quality claims could be tested here. It suits people who want a simple local chat app across many devices and are comfortable with a young project; anyone who needs answers at the level of large cloud models or chat history across devices should wait for the planned features or choose another tool.

Sources

← Back to Power Local LLM

MCP