DeepInfra
Call hosted models through an API without managing the serving infrastructure.
Deep Infra provides an API platform for running AI model inference.

Deep Infra
Founded 2022 · Palo Alto, California, US
Official Website$107M
DeepInfra said it would expand compute capacity, improve developer tools and support more open-source models.
Led by 500 Global, Georges Harik. Participation included A.Capital Ventures, Felicis, NVIDIA.
Source · deepinfra.com ↗US$18 million
DeepInfra presented the financing as part of its work on hosting and serving open-source models.
Led by Felicis, Georges Harik.
Source · deepinfra.com ↗Nikola Borisov, Georgios Papoutsis, and Yessen Kanapin brought experience running large backend systems to DeepInfra. They started the company after seeing investment in training outpace the infrastructure needed to run models in production.
Call hosted models through an API without managing the serving infrastructure.
DeepInfra built its business around running inference, then announced plans to expand capacity internationally. Its work with NVIDIA’s open AI ecosystem added support for Nemotron models and Dynamo serving software, broadening what could run on that infrastructure.
DeepInfra offers a catalog of hosted models through an API. Baseten also serves models, with infrastructure for teams to bring and deploy their own workloads. Teams choosing between the two can start with whether a catalog endpoint meets their needs or they need to configure the serving stack.