Serve a model in production

Run a trained model behind an API at scale — hosted endpoints, GPU capacity, routing, and the cost and latency trade that comes with them. 180 products across 86 companies in the Ludbee catalogue carry this task — spanning every layer of the stack that exists here, not one of its own: 22 Application, 32 Infrastructure, 19 Foundation Models, 13 Hardware.

Filter the whole directory by this job to sort and compare these products.

Where these run

SaaS
125 of 176 products that record it
Self-hosted
52 of 176 products that record it
Private cloud
21 of 176 products that record it
Hybrid
18 of 176 products that record it
On-device
18 of 176 products that record it
Air-gapped
2 of 176 products that record it

4 of 180 products carry no recorded deployment yet, and one product can run more than one way — the counts above are of records, not of products.

What this page does not claim

This list carries no score and no recommendation, and one order only: inside each layer, a company whose product for this task is marked AI feature — sold for another job, with AI as one function its own page names — comes after the rest, A–Z within each. The standing rule that applies to the whole directory applies here too: no payment changes a figure, a ranking or the order of any list, this one included. Nor can it yet say what governs any of these products once deployed — what data they can reach, what they are certified against, whether a vendor trains on what you send it. That field is defined in the ontology and is populated on none of the 3,888 products in the catalogue yet, so a page that leaned on it would be promising an answer the data cannot back. It will say so here the day that changes.

The companies

Application (22)

Infrastructure (32)

Foundation Models (19)

Hardware (13)