Innovation & Tech

Baseten Lands in Microsoft Foundry Preview With Four Open Models

Baseten's inference infrastructure powers DeepSeek V4.1 Flash, GLM-5.3 and Kimi K3 in Microsoft Foundry preview, with 100% of spend counting toward MACC and GA set for early 2027.

By Grace Kim

3 min read

Baseten Brings Four Open Models to Microsoft Foundry Preview
Baseten Brings Four Open Models to Microsoft Foundry PreviewTho-Ge / pixabay

What's News

  • Baseten entered Microsoft Foundry public preview on November 17, 2026 with four open models: DeepSeek V4.1 Flash, GLM-5.3, GLM-5.3 Flash and Kimi K3.
  • Baseten says 100% of preview usage billed through Azure counts toward customers' Microsoft Azure Consumption Commitment (MACC).
  • The infrastructure targets 99.9% uptime; Baseten stores neither prompts nor responses.
  • General availability is expected in early 2027, with new open models added on their release day.
  • Administrators must opt in before organizations can access the previewed models.

Baseten's inference infrastructure entered public preview through Microsoft Foundry on November 17, 2026, powering four open models for Azure customers as a first-party Microsoft service. The launch lineup: DeepSeek V4.1 Flash, GLM-5.3, GLM-5.3 Flash and Kimi K3.

The models sit in the Foundry catalog alongside other open and closed systems, under the same enterprise controls and customers' existing Microsoft agreements. All Foundry customers can use the preview — but an administrator must opt in before an organization gets access. Baseten names coding agents, research workflows and customer support applications among the target uses. The deal's immediate value for Azure teams is operational, not experimental.

What does the integration actually change?

Baseten positions the integration as a way to evaluate and adopt open models without sourcing GPU capacity or managing model runtimes. Baseten handles the serving infrastructure. Developers pick between two deployment types:

  • Standard deployments — Global Standard and Data Zone Standard in the US — use pay-per-token billing.
  • Provisioned deployments — Global Provisioned and Data Zone Provisioned in the US — use provisioned throughput units (PTUs), which reserve capacity and deliver predictable throughput.

Both deployment paths are ready in seconds, according to the company. Speed of provisioning, not raw model selection, is the practical hook here.

How do developers connect?

Models are served through the Chat Completions API and the Responses API. Baseten says that support lets existing OpenAI-compatible code and agent workflows carry over with minimal changes. That cuts the integration work needed to test the new catalog entries — teams already writing against OpenAI-style endpoints can point them at Foundry instead of rebuilding.

What about security, billing and infrastructure?

The service runs on the same performance research and inference infrastructure behind Baseten's Model APIs. Foundry contributes the Azure security, observability and billing framework. Baseten says every model is tuned across routing, batching, caching and kernels, and the infrastructure targets 99.9% uptime.

The data-handling posture is strict: requests are stateless, and Baseten stores neither prompts nor responses. Authentication runs through Foundry with Microsoft Entra ID and key-based access, while Azure Policy supplies per-model access controls.

The billing detail matters most for procurement teams. Usage is billed through Azure as first-party Foundry consumption, and Baseten says 100% of that spending counts toward a customer's Microsoft Azure Consumption Commitment (MACC). Model usage now ties directly into an existing Azure spending arrangement rather than a separate vendor line item.

Steve Sweetman, VP of Product at Microsoft Foundry, said: "Bringing Baseten to Microsoft Foundry gives enterprises more choice and pairs Baseten's optimized and reliable inference with the governance, controls and observability within the Foundry platform so developers can turn their most ambitious ideas into production applications."

When does general availability arrive?

Public preview is phase one. Baseten expects general availability in early 2027 and plans to keep adding open models as they're released — including making new systems available on their release day. That day-one cadence would put Foundry in direct competition with independent inference providers on freshness of open-model access.

For Azure teams, the near-term math is straightforward: open models can now be evaluated through Foundry's catalog, controls and billing without building a GPU serving environment. The preview is a working test of whether integrated infrastructure and procurement can shorten the path from model evaluation to production for enterprise AI teams.

Source: unite.ai

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.

2,301 articles

Related articles

« Previous articleNext article »