To show our commitment to this project, we’ve Just locked the 34.2M ZKME tokens bought from pumpfun, during launch with
@Streamflow_Fi
https://t.co/P8zGLAkEcY
you shouldn't need to know what an "A100-hour" costs to rent compute
just say it:
"train a small model for ~4 hours"
the zKmex payment agent turns plain language into a job spec + SOL quote. read it, confirm, done.
most compute marketplaces have one problem: GPUs sit idle between jobs, earning $0.
zKmex fixes the dead time. idle cards mine a base SOL yield automatically. the second a renter shows up, they switch to the job.
your hardware never sleeps
how zKmex prices compute:
base $/GPU-hr by tier
· consumer (4090) → $0.40
· pro (A100) → $1.80
· datacenter (H100) → $3.50
× demand multiplier (rises with utilization)
→ converted to SOL at live price
busier the network, higher the rate. providers earn more exactly when demand is real
"how do i know the job actually ran on the GPU i paid for?"
every completed job on zKmex ships a work receipt:
sha256(job ‖ node ‖ spec ‖ result ‖ nonce ‖ salt)
you verify it yourself. provider can't get paid for work that doesn't match. trust nobody, check everything
the GPU sitting in your rig right now is either making you money or collecting dust
zKmex makes it the first one
plug it in → renters run AI jobs on it → you get paid in SOL per GPU-hour. when nobody rents it, it mines a base yield instead of idling