janhq/jan

Choose this fork only if you need a near-vanilla Jan mirror and are comfortable being 209 commits behind upstream. For most adopters, upstream is the better choice because it is actively maintained and this fork adds no visible features.

Choose this fork only if its engine-focused changes are exactly what you need. For most users, upstream Jan is the safer default because this fork is materially stale and likely missing major recent product improvements.

Choose this fork only if you want a pinned, offline-focused Jan variant and are prepared to live without current upstream progress. For most adopters, upstream is the safer choice because this fork is stale and materially diverged.

Prefer upstream unless you specifically need the fork’s custom docs, provider fixes, or model-handling changes and are prepared to maintain a substantial delta. For most adopters, the maintenance and compatibility risk outweigh the local gains.

Choose this fork if your goal is a customized local-inference product with multiple engine support and you can absorb maintenance debt. Choose upstream Jan if you want an actively maintained, broader-feature desktop client with current fixes and better long-term support.

Choose this fork only if you specifically need its older, engine-focused direction and are willing to own long-term maintenance. For most adopters, upstream Jan is the safer default because this fork is materially stale and highly divergent.

Prefer upstream unless you specifically need this fork's customizations or are prepared to own a large rebase burden. This fork looks like a stale, highly divergent maintenance branch, not a drop-in upgrade path.

Choose this fork only if you specifically need its custom proxy, CLI, or plugin-runtime changes and are prepared to maintain a large, stale divergence. For most adopters, upstream Jan is the safer choice because it is much newer, actively maintained, and likely more complete.

Prefer upstream unless you intentionally want an unchanged historical snapshot. This fork adds no visible capabilities and is far enough behind that it is missing substantial recent upstream work.