MeshCore vs Meshtastic vs Reticulum

First: they’re not three of the same thing

Almost every bad comparison starts here. Meshtastic and MeshCore are the same kind of thing — an integrated product where you flash firmware onto a cheap LoRa board, pair a phone, and send messages. They genuinely compete.

Reticulum is a different layer of the stack. It’s a networking stack, not a messaging app and not a radio firmware — closer in ambition to “an alternative to TCP/IP” than to “an off-grid texting app.” Install it and you get a daemon and a library; there is nothing to chat with until you add an application on top (Sideband, MeshChat or NomadNet). It runs over LoRa, but also over WiFi, TCP/IP, packet radio and serial — and it will route a single logical network across all of them at once.

The one-line version. Meshtastic and MeshCore ask “how do we build the best off-grid messenger?” Reticulum asks “how do we build a network that doesn’t need the internet?” Comparing the third to the first two on messaging features misses most of what it is.

Versions move fast. Everything here reflects July 2026: Meshtastic firmware 2.7.x, MeshCore firmware 1.16.0, Reticulum (RNS) 1.4.0.

Meshtastic uses managed flooding. A node that hears a packet waits a short random moment and rebroadcasts it — unless it hears someone else relay it first, in which case it stays quiet. The wait is weighted by signal strength so the weakest-heard (farthest) node transmits first, pushing the packet outward. It’s bounded by a hop limit: default 3, maximum 7. Since firmware 2.6, direct messages don’t work this way any more — a DM floods once to find the recipient, then subsequent messages are relayed by a single learned next-hop node, falling back to flooding only if that hop goes quiet.

MeshCore floods to discover, then routes. The first message floods; the destination reverses the path that accumulated in the packet and sends it back; the sender caches it and every later message carries that explicit path, so only the listed repeaters retransmit. But MeshCore floods too — node adverts, every new conversation’s path discovery, and all public and group channel traffic, which by definition has no single destination.

So the real difference isn’t “flooding vs routing.” Both flood, and both route. The difference is who is allowed to repeat. In MeshCore, your handset never relays — only nodes flashed as repeaters or room servers do. In Meshtastic, every node relays unless you configure it not to. MeshCore separates users from infrastructure by construction; Meshtastic separates them by configuration discipline — and its own current guidance is that most nodes should be set to a non-relaying client role.

Reticulum does neither. Destinations periodically announce themselves; transport nodes remember only the direction one hop closer, never a whole path. Data packets are never flooded — they’re handed hop to hop along those learned next hops, across whatever mix of LoRa, WiFi and internet links exists.

An honest caveat on scaling. It’s tempting to declare a winner here. Neither project publishes a benchmark against the other, so treat “X scales better” as an architectural expectation, not a measured result. What is documented: Meshtastic congestion in dense metros is real, and the project attributes it mostly to airtime — slow presets plus chatty position and telemetry broadcasts — rather than to flooding itself. It auto-stretches broadcast intervals past 40 nodes and suggests moving off the slowest preset past roughly 60.

No — not in any pairing, and not by accident. They’re separate protocols with different framing, addressing and cryptography. A node on one network cannot exchange messages over the air with the other two, and there’s no channel name or region setting that bridges them. It’s deliberate: when someone proposed MeshCore repeaters also relay Meshtastic packets, the answer was a plain “no interoperability planned.”

They also don’t collide as much as you’d think, because the defaults are separated: US Meshtastic sits at 906.875 MHz on a wide, slow preset, while MeshCore’s US preset is 910.525 MHz on a much narrower one. Different LoRa sync words also mean Meshtastic won’t even demodulate the others’ frames.

But don’t mistake that for immunity. A sync word stops a radio decoding a packet; it does nothing to stop the energy. A 22 dBm transmitter a metre from your receiver will trash reception no matter whose firmware it runs. Co-located nodes need frequency separation, antenna spacing and sometimes filtering — ordinary RF site engineering that applies identically to all three.

Same hardware, one at a time. All three run on the same class of boards — Heltec V3, RAK4631, T-Beam, T-Deck. But a board runs one firmware, and its single radio serves one protocol. There is no dual-stack mode; switching means reflashing, and being present on two networks means two radios.

Bridges exist, with caveats. Host-side software can relay between ecosystems, and Reticulum can even be tunnelled over a Meshtastic or MeshCore mesh (using it as a dumb link layer). None of this is native radio interop: it needs an extra computer and usually a second radio, it isn’t shipped by the projects, and critically — a bridge terminates end-to-end encryption at the bridge. Whoever runs that machine sees everything crossing it.

Start with what’s already on the air near you. A mesh with no peers is a paperweight. Whichever stack your local group actually runs will beat a technically superior stack with nobody on it — and that single consideration outweighs every architectural argument above. Around here, that’s MeshCore, which is why most of our guides lean that way. Check your own area before you buy anything.

Pick Meshtastic if you want the shortest path from unboxing to a working message; you’re kitting out hikers, a race crew or a family; your nodes move; or you want the widest hardware choice, the deepest pool of tutorials, and a stack that’s open source top to bottom. Its self-healing behaviour genuinely suits unplanned, mobile deployments. Accept that dense city meshes get congested — and that being a good neighbour means running current firmware, turning telemetry down, and setting most of your nodes so they don’t relay.

Pick MeshCore if you’re building deliberate infrastructure: fixed repeaters on hills and rooftops, serving a town, with a group that will plan placement. Because handsets never repeat, adding users doesn’t add congestion — the property that matters most as a network grows. Its delivery reports and path visibility also make it pleasantly diagnosable. Accept that it’s repeater-dependent (no repeaters, no reach), that it handles mobility less gracefully because learned paths break, and that the official apps are proprietary.

Pick Reticulum if you want a network rather than a messenger: linking a LoRa segment to WiFi, to a tunnel over the internet, to packet radio, all as one addressable whole — with strong anonymity and forward secrecy as defaults, plus file transfer, pages, even voice. Being a user is easy now; being an operator is still meaningfully harder than flashing a board and opening an app, with thinner documentation and fewer people to ask.

Or just run two. Nodes are $20–50. Running a second stack costs less than most people’s antenna budget, and plenty of operators keep a MeshCore repeater and a Meshtastic node on the same mast. “Both” is a perfectly good answer.

None of these is a life-safety system. All three are unlicensed-band, best-effort, hobbyist-maintained and direction-findable. They’re superb for community, resilience and fun — but if you’re planning around emergencies, build in redundancy and don’t let any of them replace licensed radio.