Monetize your
idle GPUs.
Keln sends traffic to your capacity and handles billing, sales, and support. Start earning within the hour.
Revenue on your terms.
Keln sends traffic to your capacity and takes the commercial work off your plate: billing, sales, marketing and customer support.
You never see whose traffic it is, and the users never see your brand.
Cap utilization, reserve GPUs, and pause any time. Your own workloads always come first.
Track earnings in real time and withdraw on your schedule: ACH, instant to debit, or wire, via Stripe.
Keln benchmarks every node on weights, capabilities, and latency. Nodes that pass get traffic, and the fastest get the most.
gpt-oss-120b$0.09 / $0.30
Qwen3.5-397B$0.12 / $0.28
Kimi-K2.6$0.95 / $4.00
GLM-5.1$0.20 / $0.55A fixed, transparent rate per token, per model. Traffic goes to the fastest node, not the cheapest.
From spare GPUs to
paid traffic in an hour.
Questions before
you connect.
Everything you need to know about serving on Keln. Still unsure?
Reach out anytime.
Keln sells inference to developers and businesses, and routes those requests to independent GPU fleets. You connect your nodes, Keln fingerprints and benchmarks them, and admitted nodes join the routing pool. Traffic is white-labeled in both directions, and Keln handles billing, support, and the customer relationship.
Yes, that is the point. Sign up, e-sign the provider agreement, and run one command on each box. The agent registers your GPUs and the models they can serve. Set a utilization cap so Keln only ever fills the headroom you offer, and your own workloads always come first.
You are paid per token at a fixed, published rate for each model, so you never have to undercut anyone to win traffic. Earnings accrue in real time and you withdraw on your own schedule.
No. Traffic is fully white-labeled in both directions. You never see whose requests you serve, and their end users never see you. Billing and support stay with Keln.
Automatically. Once you e-sign the provider agreement you can connect a node immediately. Most providers are serving paid traffic within an hour.
Yes. Set a utilization cap, reserve specific GPUs, schedule maintenance windows, or pause entirely. Keln only ever fills the headroom you offer.
Only what's needed to serve the request, in transit. Nothing is persisted on your side: zero data retention is enforced platform-wide.
Keln keeps a small margin on each model it serves. You see your rate per token up front, before you connect. Details in the Provider docs ↗
Earnings accrue in real time. Withdraw via ACH, instant-to-debit, or wire, settled through Stripe on the schedule you choose.