Microsoft Copilot Adds GPT-6.1 Sol and Claude Sonnet 5.5 Across Microsoft 365

The addition of GPT-6.1 Sol and Claude Sonnet 5.5 gives eligible Microsoft Copilot users more flexibility when choosing the AI models they want to use for their work.

Instead of relying on a single model for every task, users will be able to select from supported options within compatible Copilot experiences. This may help organizations explore different approaches to writing, research, data analysis, and content creation without leaving their existing Microsoft applications.

The announcement also reflects a broader shift in enterprise software. Businesses increasingly want access to multiple AI models within familiar productivity tools, alongside the administrative controls needed to manage usage, costs, and compliance.

However, the availability of additional models does not automatically mean that every user will see them immediately. Microsoft is implementing the change in phases, and access depends on the relevant subscription and configuration requirements.

Which Microsoft Copilot Products Are Getting the New Models?

The rollout covers six major product areas, with different availability schedules and licensing arrangements.

Microsoft Word: GPT-6.1 Sol and Claude Sonnet 5.5 are scheduled for phased availability.
Microsoft Excel: Eligible users will gain additional model choices for supported spreadsheet-related tasks.
Microsoft PowerPoint: The new models are being introduced as part of the wider Microsoft 365 Copilot rollout.
Microsoft 365 Copilot Chat: Users with eligible licenses will be able to select supported models as they become available.
Copilot Cowork: The rollout began September 30, 2026, with usage-based billing.
Microsoft Copilot Studio: The rollout also began September 30, with usage-based billing.

According to the announcement, Word, Excel, PowerPoint, and Copilot Chat were scheduled to begin receiving the models in the following week.

The distinction between the initial rollout and the phased deployment matters for organizations planning internal communications. Even when a feature has been announced, availability can differ between users and tenants.

Administrators should therefore verify access in their own environments before promising a specific launch date to employees.

What GPT-6.1 Sol and Claude Sonnet 5.5 Mean for Users

The main benefit of the update is greater choice. Eligible users will be able to select GPT-6.1 Sol or Claude Sonnet 5.5 from the model selector in supported Copilot experiences, provided their organization’s configuration allows access.

For employees, this creates an opportunity to experiment with different models when completing routine work. A user might compare outputs when drafting a report, summarizing information, analyzing a spreadsheet, or preparing a presentation.

However, the announcement does not establish that either model consistently outperforms the other. It also does not provide enough comparative evidence to conclude that one is the best option for every business task.

Organizations should evaluate the available models against their own needs rather than assuming that a newer model will automatically produce better results.

Important evaluation criteria include:

Accuracy: Does the model provide reliable answers based on the available information?
Reasoning: Can it handle complex instructions and explain its conclusions clearly?
Productivity: Does it reduce the time needed to complete routine tasks?
Consistency: Does it produce dependable results across repeated requests?
Governance: Does its use align with the organization’s security, privacy, and compliance requirements?

Testing representative business tasks can help companies determine whether additional model choices translate into meaningful improvements.

Usage-Based Billing and Limits: What Businesses Should Know

The announcement distinguishes between Copilot Cowork and Copilot Studio, which use usage-based billing for this rollout, and Microsoft 365 applications covered by the applicable user subscription license.

For Cowork and Copilot Studio, organizations need the appropriate user subscription license and usage-based billing enabled. Administrators should understand the relevant consumption arrangements before making the new models widely available.

In Word, Excel, PowerPoint, and Copilot Chat, the models are being introduced under the Microsoft 365 Copilot user subscription license. Usage limits still apply.

Microsoft says these limits are designed to support broad day-to-day AI use across organizations and that most users are not expected to reach them. Users approaching a limit will receive notifications within the product.

If a user reaches a limit, they can continue working by switching to Auto mode or selecting another available model.

This fallback is important because reaching a model-specific limit does not necessarily mean that Copilot becomes unusable. Employees may still be able to complete their work through other supported options.

For IT and finance teams, the priorities are to understand billing arrangements, communicate applicable usage policies, and monitor adoption after the rollout.

Administrator Settings Determine Access

Although the feature is enabled by default for eligible users who meet the applicable requirements, organizations may need to review their AI provider settings before both models become available.

Access to Claude Sonnet 5.5 requires the Anthropic subprocessor to be enabled in the organization. GPT-6.1 Sol may require the OpenAI subprocessor to be enabled in certain Copilot experiences.

Microsoft also distinguishes between OpenAI-operated models and models operated by Microsoft through Azure OpenAI. Disabling the OpenAI subprocessor affects the relevant OpenAI-operated models, but it does not disable Microsoft’s Azure OpenAI models.

These distinctions are important for administrators managing enterprise AI access.

Before communicating availability to employees, organizations should:

Confirm that users have the required licenses.
Review the Anthropic subprocessor setting if Claude Sonnet 5.5 access is planned.
Review the OpenAI subprocessor setting where GPT-6.1 Sol requires it.
Check whether additional product-specific configuration is necessary.
Verify which models appear in the supported Copilot experiences.
Prepare helpdesk teams to answer questions about model availability and usage limits.

These steps can help reduce confusion when employees see different options across applications or accounts.

Security, Privacy, and Compliance Considerations

Adding third-party AI models to enterprise productivity software creates new opportunities, but it also makes governance an important part of deployment planning.

Organizations should review the data-processing arrangements and contractual commitments applicable to their chosen models. They should not assume that every model has identical data-handling, residency, or compliance characteristics.

For companies operating in regulated industries, these details can influence whether a particular model is appropriate for specific workflows.

IT, security, and compliance teams should consider the following questions:

Does the model’s processing arrangement meet the organization’s contractual obligations?
Are applicable data residency requirements satisfied?
Which employees and departments should have access?
Do existing AI usage policies cover the new model options?
Should sensitive business information be restricted to particular workflows or approved configurations?

The goal is to ensure that greater model flexibility does not create uncertainty about how employees should use AI.

A review of existing policies, access permissions, and data-processing requirements can help organizations introduce the new options responsibly.

How Businesses Can Prepare for the Rollout

Organizations do not need to take action simply to enable the feature for eligible users. However, administrators can take several practical steps to make the rollout smoother.

1. Check availability and licensing

Confirm that employees have the required subscriptions and review the applicable provider settings. Check the model selector in the relevant applications to determine whether access has reached your tenant.

2. Communicate the changes

Explain which models may become available, how users can select them, and what to do if a usage limit appears.

Clear communication can reduce unnecessary support requests and help employees understand their options.

3. Evaluate the models against real tasks

Choose representative activities, such as summarizing reports, drafting internal documents, analyzing spreadsheet data, or preparing presentations.

Compare the results using consistent criteria, including accuracy, output quality, time saved, and the amount of human review required.

4. Review costs and governance

For products using usage-based billing, establish appropriate monitoring and internal guidance. Review model access permissions and confirm that the relevant data-processing arrangements meet organizational requirements.

This approach allows businesses to explore the new capabilities while retaining oversight of their AI deployments.

The Future of AI in Microsoft 365

Microsoft’s expansion of Copilot reflects a growing demand for AI flexibility within enterprise software. Businesses increasingly want employees to access different models through familiar applications while keeping licensing, administrative controls, and security requirements in view.

For employees, the ability to choose between supported models could provide more options for tackling everyday tasks. For administrators, it introduces additional decisions about access, cost management, and governance.

The real measure of success will be whether those choices improve business outcomes. Organizations will need to assess which models work best for their workflows, how usage affects costs, and whether their existing policies adequately address the expanded selection.

As the rollout continues, businesses should monitor availability in their own environments and evaluate the models using practical, measurable criteria. The organizations best positioned to benefit will be those that combine experimentation with clear policies, appropriate oversight, and a focus on real productivity gains.