Baseten

Baseten is a model inference infrastructure platform for developers, startups, and enterprises that need to deploy, serve, and scale open-source, fine-tuned, and custom AI models in production.

Key capabilities include:

• Two-click deployments of open-source models from a hosted model library for fast prototyping and production evaluation.
• Custom and fine-tuned model serving using Truss to package models for production deployment.
• High-performance inference at scale with distributed multi-cloud infrastructure, including dedicated deployments and cross-cloud high availability.
• Support for multiple modalities (LLMs, embeddings, image generation, transcription, and text-to-speech) with production-focused runtimes and developer workflows.

Pricing: Unknown (pricing page available on the website).

Notable for its emphasis on cost-performance and inference optimization, including applied performance research and integration with common inference frameworks such as NVIDIA Dynamo, TensorRT-LLM, SGLang, and vLLM, plus partnerships with Google Cloud and NVIDIA. Founded in 2020.