Deploy a model on Lambda without Lambda becoming the product.

Lambda is a GPU cloud in the denv catalog. Status: available. You connect with your own Lambda API key. denv launches and terminates machines in your account. Lambda bills you.

GPUs denv has listed: A10, A100 40GB, A100 80GB, H100, B200. Regions denv has checked: US, Europe, Asia and the Middle East (13 regions). What you can launch still depends on your account and on Lambda’s capacity that day.

What denv uses Lambda for

A site. The site has a GPU type and a minimum and maximum number of machines. A recipe runs on that GPU type. The endpoint can also include Nebius, EC2, Hetzner or your own servers. If Lambda is out of the GPU you asked for, failover only helps if another site is already allowed and healthy.

H100 and B200 have their own pages. There is a public H100 measurement. There is not a public Lambda-versus-anyone price, and denv will not invent one. You pay Lambda.

Driver install, MIG and DRA, and repair run on Lambda nodes the same way they run anywhere else the node can report GPU faults.

The product is early. See the catalog and the quickstart.