Ente, the encrypted storage company behind the privacy-focused Ente Photos service, released Ensu on March 2, 2026 — an open-source application that runs large language models entirely on a user’s local device. No internet connection is required, no queries are transmitted to external servers, and the app is available at no cost.
- Ensu runs LLMs fully on-device; all processing is local with zero data sent externally
- Available for iOS, Android, macOS, Linux, Windows, and an experimental web version
- Currently an Ente Labs project; acknowledged by the company as less capable than ChatGPT or Claude
- Future updates will add end-to-end encrypted chat sync via Ente accounts or self-hosted backends
What Happened
Ente, an encrypted storage company best known for its privacy-first photo backup service, released Ensu on March 2, 2026 — a free, open-source application that runs large language models entirely on a user’s device, with no internet connection, no subscription fee, and no data sent to external servers. The announcement was published on the company’s official blog at ente.com/blog/ensu; author details were not available at time of publication.
Ensu functions as a ChatGPT-style conversational interface that operates entirely offline. A planned feature will allow users to optionally back up and sync their chat history across devices via Ente accounts or self-hosted infrastructure, with end-to-end encryption — though this capability is not yet available in the initial release.
Why It Matters
The release addresses what Ente describes as two compounding risks with centralized AI: the privacy limitations of cloud providers — including arbitrary content bans, non-portable chat history, and opaque content filtering — and the potential for centralized model control to enable large-scale manipulation of public discourse. “LLMs are too important to be left to big tech,” the company stated directly in the announcement.
Ente draws a parallel to its earlier work on Ente Photos, where the company built face recognition, person clustering, and natural language image search that all run entirely on-device. “People called us crazy,” the company wrote of that effort. That implementation is now in active daily use by Ente Photos customers, and the company cites it as a precedent for local AI being achievable despite initial skepticism.
The company also identifies structural dependency problems specific to centralized AI platforms: account bans that revoke access without recourse, memory systems that do not transfer between providers, and content shaping by platform operators. These are presented as design-level problems that on-device processing eliminates rather than mitigates.
Technical Details
Ensu supports image attachments and runs as a native app on iOS, Android, macOS, Linux, and Windows, with an additional experimental web-based version; all query processing occurs locally, and the app currently has no web search capability, making it fully functional in offline environments. The source code is publicly available for community review and contribution.
Ente frames its capability argument around a “threshold” concept rather than parity with frontier systems: the position is not that local models need to match GPT-4 or Claude, but that once they cross a utility threshold sufficient for everyday tasks, the size of the remaining gap becomes secondary. The company cited three use cases where the current version already performs well: discussing classic texts (specifically naming the Bhagavad Gita and the Bible as areas where the model demonstrates strong recall), offline conversation without connectivity, and handling topics users would not want routed through a cloud provider.
The company acknowledged the current release is “not as powerful as ChatGPT or Claude.” No benchmark data, hardware requirements, or underlying model names were disclosed in the announcement.
Who’s Affected
Privacy-conscious individuals, professionals handling sensitive information, and users in low-connectivity environments are the clearest immediate audience, though the open-source codebase and self-hostable sync layer make the project relevant to developers building or evaluating privacy-first AI tooling. The cross-platform availability — iOS, Android, macOS, Linux, and Windows — means the app is not restricted to any single device ecosystem.
Users who have encountered content restrictions, data retention concerns, or account limitations on centralized AI platforms have a direct offline alternative in Ensu, though they accept a capability trade-off relative to cloud-hosted models in the current version.
What’s Next
Ente’s stated roadmap includes encrypted chat history backup and cross-device synchronization through Ente accounts or self-hosted deployments, but no delivery timeline has been announced; the company is treating the project as a Labs-stage experiment without pricing or stability commitments, and has stated it wants to establish product direction before introducing commercial considerations.
Further capability improvements to Ensu will depend in part on continued development of open-source models compact enough to run efficiently on consumer hardware. Ente has not disclosed which underlying models the current version uses or the minimum hardware specifications required to run it.
