Baseten announced on November 17, 2026, that its open-model inference service had entered public preview through Microsoft Foundry. Baseten expects general availability in early 2027. The company’s announcement describes how Azure organizations can enable access and what the service offers.
How organizations access the preview
Baseten says an organization administrator must opt in before developers can access the service. Once access is enabled, developers can find the models in Microsoft Foundry’s catalog and choose a deployment type.
For an Azure team, the first step is organization-level opt-in; developers choose a deployment after access is enabled.
Deployment, APIs, data handling, and billing
Baseten describes two deployment types, Standard and Provisioned, and says the service supports the Chat Completions and Responses APIs. The company says teams using OpenAI-compatible code and agent workflows can carry them over with minimal changes.
Baseten also describes requests as stateless and says it stores neither prompts nor responses. It says usage is billed as first-party Foundry consumption through existing Microsoft agreements.
Baseten’s stated general-availability target is early 2027.