Claude Opus 5 Is in Microsoft 365 Copilot. Should UK Firms Switch It On?

Anthropic released Claude Opus 5 on 24 July, and Microsoft added it to the Microsoft 365 Copilot model selector the same day. It is rolling out alongside GPT-5.6 in Word, Excel, PowerPoint, Copilot Chat, Cowork and Copilot Studio.

If you are a UK firm there is a catch you have met before. Anthropic models are switched off in your tenant by default, and turning them on is a data decision before it is an IT one. We covered that briefly when GPT-5.6 arrived earlier this month. Opus 5 makes it worth revisiting properly, because the model sitting behind that switch just got considerably better.

Key takeaways

• Claude Opus 5 is rolling out to the Copilot model selector across Word, Excel, PowerPoint, Chat, Cowork and Copilot Studio.
• GPT-5.6 remains the preferred default and handles most day to day work well. Opus earns its keep on long, complex, high stakes tasks.
• Anthropic models are off by default for UK, EU and EFTA tenants because that processing happens outside the Microsoft EU Data Boundary.
• There is no extra licence cost. Anthropic models are included with your Microsoft 365 Copilot licence at no additional charge.
• Enabling it is a fifteen minute job in the admin centre. The governance conversation around it takes longer, and it should.

What actually changed

Anthropic positions Opus 5 as an everyday model for knowledge work rather than a specialist tool, with improvements in multi step reasoning, long documents and professional analysis. It also carries an effort setting that trades compute against speed and cost, which matters more than it sounds when you are running work at scale.

For Copilot users the practical change is simpler. Where Claude is enabled, Opus 5 appears in the model picker by name. Anthropic's naming carries through into Copilot, so you see Sonnet and Opus rather than a generic label. That is a small mercy, because on the OpenAI side Microsoft does not surface the GPT-5.6 tier names. OpenAI ships Sol, Terra and Luna. Copilot users generally see friendlier wording along the lines of quick response and think deeper, which tells you how much reasoning you asked for but not which model you got.

One thing to be aware of before you go looking: Microsoft's own documentation is running behind the rollout. The Learn page for choosing a Cowork model was last updated in mid June and still lists the previous generation. Trust the picker in front of you rather than the docs, and expect the available list to differ between Word, Excel, Chat and Cowork while this settles.

When to use which

Start with the honest answer. For most work, leave it on Auto. That is Microsoft's own guidance and it holds across the suite. Copilot picks a model based on the task, and on a routine email or a meeting summary you are unlikely to notice which one answered.

The picker earns its keep on specific jobs, and there the differences are real.

Reach for GPT-5.6 when

You want speed and efficiency on volume work, you need drafting with citations, or the task is well defined and you mostly want it done quickly. It is the tuned default for Microsoft 365, jointly optimised by Microsoft and OpenAI for knowledge work, and it does more with fewer tokens than the model it replaced. It also needs no opt-in and sits under Microsoft's standard Copilot commitments, which for some firms settles the question on its own.

Reach for Claude Opus 5 when

The work is complex, multi step and carries consequences. Microsoft's guidance for the Opus family points at deep reasoning, multi step analysis, complex research and writing. In practice that means the long document where structure and argument have to hold together from first page to last, the financial model where the reasoning matters more than the turnaround, the board paper, the tender response, the due diligence review.

Reach for Claude Sonnet when

You want Claude's instruction following on ordinary tasks without the wait.

There is also a paired mode in Cowork called Sonnet plus Opus Advisor, where Sonnet does the work and Opus reviews the result for accuracy and completeness. For an important deliverable that is a genuinely useful pattern, because you get a second pass without having to manage two conversations yourself.

A rule of thumb we give clients: if you would have handed it to a junior to draft, Auto is fine. If you would have asked a senior to check it, switch to Opus.

Why it is switched off in your tenant

Microsoft enables Anthropic models by default for most commercial cloud customers. The exceptions are the EU, EFTA and the UK, where the setting ships as No users.

The reason is where the processing happens. Anthropic models used in Microsoft 365 Copilot sit outside the Microsoft EU Data Boundary, and outside in country processing commitments where those apply. Microsoft's position is that the contractual protection still holds. Anthropic has been onboarded as a Microsoft subprocessor, so the Microsoft Product Terms and the Data Protection Addendum apply, Enterprise Data Protection covers the usage, and the Microsoft Customer Copyright Commitment extends to Anthropic models in Copilot and Copilot Studio. What you do not get is the geographic guarantee.

That is the trade you are being asked to make, and it belongs with your DPO or compliance lead rather than with whoever happens to be logged into the admin centre.

One thing to check if you enabled Claude at some point in the past: the old Anthropic independent processor setting was decommissioned on 1 May 2026. If your tenant opted in under Anthropic's separate commercial terms back then, that consent no longer carries anything. You have to opt in again as a subprocessor or you have no access at all.

How to switch it on

You need the AI Administrator or Global Administrator role.

1. Go to the Microsoft 365 admin centre and select Copilot, then Settings, then View all.
2. Select AI providers operating as Microsoft subprocessors.
3. Under Available subprocessors for your organization, select Anthropic and save.
4. Under Choose who can access Anthropic models, pick your users or groups and save.

Do not skip the last step. You can scope access to specific users or Entra ID security groups rather than opening it to the whole tenant, and for a first rollout you should. Pick a pilot group who actually do the complex work, measure the output quality for a few weeks, then widen. The assignment applies at provider level and is enforced consistently across Microsoft 365 Copilot and Copilot Studio.

Two settings that catch people out

The apps level setting. Since April 2026 there has been a separate control called Copilot in M365 apps with Anthropic models, covering Copilot experiences in Word, Excel and PowerPoint for UK, EU and EFTA tenants. It is on by default for tenants created after 25 March 2026. If your tenant predates that date, check the Message Center for your default. The part that trips people up is this: once you enable the main subprocessor setting for all users or for specific users and groups, this apps level setting becomes unavailable and cannot be changed.

Copilot Studio and Power Platform. Enabling Anthropic in the Microsoft 365 admin centre is necessary but not sufficient for agents. There are additional controls in the Power Platform admin centre governing external large language models for generative responses. If Claude is not appearing in your Copilot Studio model selector, that is almost always where the block is.

Worth noting too that Anthropic models are not available at all in GCC High, DoD or other sovereign clouds. Non federal GCC customers gained an opt in on 22 July 2026.

What it actually costs

Here is the part that surprises people, first pleasantly and then less so.

There is no per model charge. Anthropic models are included with the Microsoft 365 Copilot licence. Switching Opus 5 on does not add a line to your invoice, there is no premium for the better model and there is no separate Anthropic subscription to buy.

You still need the Copilot licence. That is the real cost and it has not changed. On Microsoft's UK list the enterprise add-on sits at £23.10 per user per month, and firms with up to 300 users have the cheaper Copilot Business SKU at £16.10. We went through the full picture in our post on how Copilot pricing quietly changed. Prices move and promotional rates come and go, so confirm the live figure in your admin centre before it reaches a budget.

Agents are metered separately. If your plan for Opus 5 runs through Copilot Studio, the model is free but the running is not. Copilot Studio and Cowork bill consumption in Copilot Credits, and consumption per interaction varies enormously depending on how the work is built. A simple lookup and a grounded multi step autonomous action are nowhere near the same cost. Our guide to keeping Copilot Credits under control covers how to model that before you commit.

The governance cost is the real one. Enabling Anthropic means accepting that a category of your business content is processed outside the EU Data Boundary. Depending on your sector that is a five minute conversation or a three month one. For a marketing agency it is unlikely to trouble anyone. For a firm handling client money, health data or legally privileged material it needs a documented decision, a data protection impact assessment and probably a scoped rollout rather than a tenant wide switch.

One trap worth avoiding. Anthropic's newest preview models sit under an entirely different arrangement labelled Preview models with Data Retention. For those, Anthropic acts as an independent processor under its own commercial terms rather than as a Microsoft subprocessor, and it retains inputs and outputs for a period. They are default off even when you have enabled Anthropic generally, and they need a separate explicit opt in. Do not enable them in the same sitting as Opus 5 without reading what you are agreeing to.

Before you flip the switch

A short checklist we run with clients:

• Confirm who owns the decision. This is a data protection call, not an IT one.
• Scope it to a pilot Entra group rather than all users.
• Check whether your tenant is affected by the decommissioned independent processor setting.
• Decide separately about preview models with data retention. For most UK businesses the answer is no.
• If agents are in scope, check the Power Platform admin centre controls and model your credit consumption.
• Tell your users. A model picker nobody knows about delivers nothing at all.

The honest summary

GPT-5.6 is your default and it is a good one. Claude Opus 5 is worth enabling if your people genuinely do complex work, and the enablement itself is a short admin job with a proper governance conversation wrapped around it. Your licence cost does not change either way.

What will not change either way is the return you get from Copilot overall. That is set by your data, your permissions model and your people's ability to use the thing well. If your results are disappointing today, a different model is unlikely to be the fix.

At IQIT we help UK firms get real value from Copilot, from hands-on training to working out what your model and data setup should actually look like. If you want a straight answer on whether Opus belongs in your tenant, get in touch.