Baseten

Baseten is an AI infrastructure company that provides a platform for running machine-learning models in production. Founded in 2019, the company says it was established to give builders the infrastructure, workflows, and tools needed to bring AI into products. It is headquartered in San Francisco, California.

The platform supports several stages of model development and operation. Customers can access hosted, pre-optimized models through Model APIs; deploy open-source, fine-tuned, custom, or proprietary models on dedicated infrastructure; and train models using managed GPU infrastructure. Baseten also offers a path to serve trained models on the same platform.

Baseten’s inference services are designed for production workloads that require performance, scaling, and operational management. The company describes its Inference Stack as combining optimized runtimes and model-performance work with GPU infrastructure, autoscaling, and model management. Its tooling supports workloads including language models, embeddings, image generation, transcription, speech generation, and multi-model or compound AI applications.

Customers can choose among managed cloud, self-hosted, and hybrid deployment options. Baseten describes managed deployments as globally scalable, while self-hosted deployments run in a customer’s own virtual private cloud; hybrid arrangements can combine customer-controlled infrastructure with Baseten cloud capacity. Dedicated and single-tenant options provide additional control over infrastructure and workload isolation.

The company serves AI startups, enterprises, and model labs building AI products and operating custom models. Its website features customer examples including Abridge, Descript, Gamma, OpenEvidence, Writer, Wispr, and Zed. Baseten also offers model distribution services for labs seeking production APIs and a channel to make models available to users.

Baseten says its founders experienced the difficulty of deploying models into production and created the company to address that challenge. Its stated mission is to build an inference platform for mission-critical AI workloads. The company’s work spans performance research, infrastructure that can scale across clouds and regions, and tools for deploying, observing, and managing models.

Baseten highlights hands-on help from forward-deployed engineers, who work with customers to build, optimize, and scale models from prototype to production. In a company funding announcement, Baseten described the team it seeks as smart, fun, and humble. These statements offer a documented view of the company’s customer-support approach and stated working culture.