Pay for scaling intelligence, not idle capacity

A GPU node is any compute instance with one or more GPUs that MLSrvyn routes inference traffic to. This includes your existing GPU fleet — MLSrvyn connects over your orchestration layer (Kubernetes, Slurm, or bare-metal). You don't pay separately per GPU; the tier limit is the number of distinct host nodes.

No. MLSrvyn is an optimization layer. You continue to pay your cloud provider or colocation facility directly for GPU compute. MLSrvyn's fee covers the traffic profiling, scaling policy engine, SLO enforcement, and bin-packing logic — not the underlying hardware.

14 days, full access to your selected tier, no credit card required. We'll help you connect MLSrvyn to your first model endpoint in under 30 minutes. If you don't see measurable latency improvement in the first week, we'll tell you exactly why and help you fix it — or the trial extends at no charge.

Yes. Upgrades are prorated to the day. Downgrades take effect at the next billing cycle. Enterprise tier pricing is negotiated annually with monthly billing options available.