Baseten

Model serving as an operational service

Baseten provides infrastructure for model inference and customization. Its work includes serving, model optimization, post-training and capacity management. The company describes tuning kernels, quantization, batching, routing and hardware selection for particular workloads, connecting model execution to the infrastructure needed to handle production traffic. [1]

Baseten's Chains tooling also supports pipelines with multiple steps, including deployment, invocation and streaming of linked components. This matters when an application needs more than a single model call: the surrounding system must connect and operate those stages as a service. The documentation describes the mechanism; it does not establish a historical launch date for the feature. [3]

A distribution route for Mercury 3

Inception's October 19, 2026 Mercury 3 release explicitly named Baseten as an access route, alongside Inception's API, OpenRouter and Vercel. Baseten's participation here is serving and distribution. Inception remains the developer of the diffusion language model, and its reported speed figures are model-vendor measurements rather than an independent comparison of inference providers. [2]