Microsoft Foundry hits 11,000 models including Claude Opus 4.8 in Excel Agent Mode

[Ellie22]

Microsoft confirmed at Build 2026 this month that its Foundry model catalogue now contains over 11,000 models, covering everything from OpenAI GPT-5.5 and Claude Opus 4.8 to Google Gemini and thousands of open source and specialised alternatives, all accessible through a single Azure endpoint. The practical implication is that enterprise developers can switch between models without changing infrastructure. Claude Opus 4.8 is also now available in Excel Agent Mode as part of the Microsoft 365 integration.

The Excel integration is more significant than it might sound for enterprise users. Putting a frontier model directly into a spreadsheet tool that billions of people use for critical business data changes the access pattern entirely. You do not need to know how to use an AI platform or write prompts in a chat interface. The model is there when you need it in a context you already understand.

The EU AI Act enforcement deadline arrives in approximately 55 days. Microsoft's multi-model catalogue strategy means they are less exposed to any single model failing compliance requirements in specific jurisdictions. Having 11,000 models to route through gives you considerable flexibility when navigating regional AI regulation.

[Ann]

The Excel Agent Mode integration is going to change enterprise AI adoption more than any standalone AI tool. Most of the people in large organisations who could benefit from AI are never going to open a separate app. Meeting them in Excel is the right move.

[Frost Jay]

11,000 models is a number designed to impress procurement teams rather than developers. The models developers actually choose for serious work number somewhere in the dozens. The catalogue size is a sales argument, not a technical one.

[Josh_79]

I have tried Claude in Excel and it is genuinely useful for the data transformation and analysis tasks that eat hours of my week. Not magic but meaningfully faster for real work.

[CMPunk]

The Azure endpoint abstraction is actually useful if you are building internal tooling. Being able to A/B test different models without infrastructure changes is worth something.

[Chris27]

The EU AI Act enforcement starting in August is going to cause a lot of quiet reshuffling. There are categories of high-risk AI use in enterprise software that are going to need compliance documentation that does not currently exist.

[Dave]

Microsoft betting on being the multimodel platform rather than owning a model is a defensible strategy. They are less exposed to any single lab's fortunes and positioned as neutral infrastructure.

[Sophie92]

Claude in Excel making sense is also a reminder that spreadsheets are still how an enormous amount of the world actually processes information. All the modern data stack tooling exists and yet Excel is everywhere.

[NightHarbour52]

The 55 days to EU AI Act enforcement should be getting more attention in tech circles. The Act has teeth and the enterprise AI use cases that are just being built now may not be compliant.