Run a model on Nebius, in your own project.

Nebius AI Cloud is in the catalog. Status: available. You connect with your own Nebius service account key. denv launches and terminates machines in your project. Nebius bills you. denv does not resell the GPU.

GPUs listed: L40S, H100, H200, B200. Regions denv has checked: eu-north1, eu-west1, us-central1.

The 50.8 tok/s starting point was one H100 in a Nebius account, called with the OpenAI client. That is one machine on one night, not a Nebius SLA and not a denv SLA.

H200 is listed here and not on the Lambda or EC2 rows. If you came from “rent H200”, the rent is between you and Nebius. denv serves the model. See H200 and H100.

A Nebius site can sit in the same cluster as any other site. Failover off Nebius requires a replica somewhere else. Repair on a Nebius node can replace the machine, because denv can launch one, up to the maximum you set.

Write to anup@denv.ai if the regions you need are not in the list above. Unlisted regions are not a quiet yes.