LobeHub vs NextChat

Last reviewed Aug 26, 2026 · 571 words

NextChat runs in 512 MB; LobeChat asks for 2 GB, and a proper multi-user install adds a Postgres database on top. That fourfold gap is the whole comparison in one number. These are two TypeScript chat front ends from the same 2023 wave — NextChat under MIT, LobeChat under Apache-2.0, both rated Easy, both dependent on provider API keys because neither runs a model — and what separates them is how much application you want wrapped around the chat box.

LobeChat is building a platform

The weight buys real things: multi-user accounts with OAuth sign-in, an assistant marketplace, a plugin system, voice conversation, vision input, and a knowledge base for chatting over uploaded documents. LobeChat wants to be the shared instance where each family member or teammate logs in, picks assistants, and keeps their own history — which is precisely why the server deployment needs Postgres, since per-user state has to live somewhere durable. Its 81,994 GitHub stars almost undersell how much product is in the box. The flip side: you are now operating a small stack, with a database to back up, credentials to manage, and more release notes to read.

NextChat is refusing to become one

Formerly ChatGPT-Next-Web, NextChat (88,653 stars) stays deliberately small. It deploys as one Docker container or with a one-click Vercel button, keeps its state client-side, and ships desktop and mobile builds so the same assistant follows you across devices. It is single-user by intent, not omission: no roles, no marketplace, prompt templates instead of a plugin ecosystem. Multi-provider support survives the diet, so switching between OpenAI, Anthropic, or a local endpoint is one settings change. On the kind of 512 MB VPS or free-tier host that LobeChat's server mode would flatten, NextChat idles comfortably. Maintenance is close to zero — the classic NextChat install is an instance nobody has touched in six months that still works exactly as it did on day one.

Both stop where the model starts

Neither app performs inference, so self-hosting either one does not by itself keep your prompts private — conversations still go to whichever provider holds your API key. Both can point at a local backend such as Ollama instead, but if local models are your main event, Open WebUI was built around that pairing and deserves the first audition — the LobeChat vs Open WebUI write-up covers that fork. Pick between LobeChat and NextChat only once the API-key workflow is the accepted premise rather than the problem.

Decision table

What I'd do

Match the tool to the head count. Just me: NextChat, deployed in ten minutes, costing nearly nothing, doing the one job fast. My household: LobeChat, because separate accounts and the assistant marketplace are what make a shared instance worth having. The mistake I see most is the middle path — running LobeChat's full server stack for a single user, which amounts to operating Postgres to store one person's chat history. Do not do that; start with NextChat and upgrade the day someone else asks for a login.