In the GitHub issues you'll consistently find reports of high memory usage in the Clash Verge Rev / verge-mihomo process: after a few hours to a day or two of running, memory climbs from a few hundred MB to several GB. Not everyone hits it, but users with many subscriptions and frequent connections are more likely to. Here are the common triggers and quick workarounds.
Common triggers
- Many subscriptions with frequent auto-update: Lots of subscriptions plus a short update interval means repeated config reloads that accumulate overhead.
- Many failed connections: When nodes can't connect, the client retries repeatedly; if some failed-connection session resources aren't released promptly, they pile up over time.
- Long uptime without restart: The longer it runs continuously, the more obvious the growth.
Things you can do right now
- Delete subscriptions you don't use; keep only the one or two you rely on.
- Lengthen the auto-update interval — e.g. set it to 1440 (24 hours) instead of something very short, so it doesn't reload constantly.
- When memory balloons, simply restart the app to free it immediately.
- Pick nodes that connect reliably to cut down on failed retries — see testing and choosing nodes.
It's tied to core session cleanup
A lot of this happens in the underlying Mihomo core rather than the UI — failed-connection sessions not being closed properly is a recurring theme in community discussion, and upstream Mihomo has had related fix commits. So keeping the client and core updated matters; see the Mihomo core upgrade notes.
If your memory use is genuinely abnormal (tens of GB), it's worth searching GitHub issues for reports on the same hardware and version, and reporting back with logs — that's more useful than waiting it out.
For node connectivity issues see subscription shows no nodes, and check the FAQ for other questions. To move to a newer, more thoroughly patched build, grab the latest installer from the download center.