hM Desktop
The app that is your own bank. Bitcoin, Lightning, dollars and private on-device AI — your money and your work, on your machine.
Burgundy looks good on you
Download v0.2.4 · Apple Silicon
Download for Mac · v0.1.2 alpha
Deprecated · previous build
On iPhone? higherMark for iOS →
Your own bank.
On your own machine.
Drive across America and it's the same skyline everywhere — fast food on every corner, bank towers over every downtown. The whole landscape is built to address you as a consumer.
hM Desktop is built for the other role. Your keys, your dollars, your AI — running locally, teetering on a new future where you hold the controls. This is our flagship, and we increasingly think it's inevitable.
Held by you
A Grid Global Account, Strike API, Arkade's VTXOs with seamless integration and increasing support across the lightning ecosystem providing you with a complementary tool for the onramps and offramps suited for the modern world... Connect to all the apps you already use, gain as much control as possible with increasing network effects.
Lightning native
Instant cash finality, send to and from anywhere interoperably on a network that already reaches millions of terminals in often just one literal second transaction time. Any second, any hour, any day of the week send to any friends, family, start hosting your own connections for them, and transact with any merchant (or become a merchant)... Lightning settles dollars instantly, you do not have to deal with any volatility, once you have funded a wallet you can send $1 and receive exactly $1 to and from anywhere and anybody in the world.
Private, on device
The model runs locally by default. Nothing leaves the machine unless you send it, and no central API server ever sees your work.
Native, not a webview
A Rust core with a GPU-drawn interface. It launches cold in under a second, and it feels like it.
One app: your keys, your dollars, your model — all of it on hardware you own.
Fast when you're on.
Still yours when you're not.
hM Desktop ships with one switch. Leave hybrid on and the heavy asks route out to Cerebras, where wafer-scale silicon has carried us from roughly 250 tokens/second, to 1,500, to — in some runs — past 2,000. That's the fastest online generation we can put in front of you.
Toggle it off and inference happens entirely on your machine. Nothing is sent, nothing is logged, and no central API server ever sees a word of it.
Local isn't the consolation prize. Our fastest test models clear 50 tokens/second on Apple Silicon and have surpassed 200 on Desktop — on par with the online interfaces most people already use every day. The difference is that this one keeps working in a storm, through a power outage, on a plane, in a hotel with a dead connection, or any time you'd simply rather nothing left the room.
This is an early alpha and the build is unsigned, so macOS Gatekeeper will refuse a normal double-click. Open it the long way once and macOS trusts it from then on.
xattr -cr /Applications/higherMark.appopen /Applications/higherMark.app
Want to test? Join the Discord or contact us and help shape it as an early user — thank you. Check back for the public release.