June 16, 2026
BlogMicrosoft’s Copilot Cowork Gets Model Choice Plus Usage-Based BillingÂ
Microsoft’s Copilot 365’s Cowork feature, which is generally available worldwide as of June 16, will offer customers a choice of models and won’t be solely Anthropic-based as it is currently. Microsoft also is adding usage-based pricing to Cowork, on top of the Microsoft 365 Copilot license charges.
Microsoft made Copilot Cowork available to Frontier testing program customers in late March, 2026. Cowork, which allows customers to run complex, multi-tool tasks, supports Anthropic’s Opus 4.8 and Sonnet 4.6 today. For customers in the Frontier test program, there’s also an option to use OpenAI’s GPT 5.5. And “coming soon,” Microsoft plans to release an Azure-hosted open-source model for Cowork that will “handle tasks at a substantially lower cost,” officials said.
(I’m assuming this is a model built by Microsoft AI, and have asked, but no word back yet.) Microsoft officials said they will release the name of the model once it’s ready for testing.
Microsoft CEO Satya Nadella spilled the beans on Microsoft’s plans to bring model choice to Cowork during an interview at Microsoft Build in early June. During that interview he said Cowork was a type of “form factor” for agents and it would be getting support for GPT in addition to Anthropic’s models. Nadella said the same way Microsoft 365 Copilot currently offers customers a choice of models, the same will be true of Cowork.
Copilot Cowork requires a paid Microsoft 365 Copilot user license, which retails for $30 USD per user per month. On top of that, Microsoft also will bill users for Cowork on a usage-based basis, with charges determined by the tasks they run. Users will be charged in Copilot Credits, with the price of each task calculated via four inputs: Model use, context retrieval, tool calls, and runtime.
There will be two payment options for Cowork: Pay As You Go (PAYG) and P3. P3 is for those who have a handle on cost projections and can commit to a usage volume in advance in exchange for a discount. Where multiple models are available, customers can use the model picker to manage cost-per-task, according to Microsoft.
Low Cost Provider
Microsoft’s strategy seemingly is to position its own models as the lower-cost option across its Copilot family as it evolves.
Microsoft is adding support for more models in its GitHub Copilot, as well. It is integrating its just-announced MAI-Code-1-Flash model to GitHub Copilot individual users and making it the default in the Auto picker there. (Currently, the default is OpenAI’s GPT 4.1.) Its plan is to continue to allow developers to select different models if they choose to do so. As of June 1, Microsoft is charging GitHub Copilot users on a usage basis.
Microsoft also is planning to release a Copilot “Super App” this summer, officials have said. That Super App will provide users with access to Copilot Chat, Copilot Cowork, Copilot Code, and Copilot Autopilots, which are agents that will run without user intervention (like its recently announced Scout personal agent, built on OpenClaw).Â
In related news, WorkIQ also is generally available as of today, June 16. And it also is adding consumption-based pricing to its model.
Work IQ is Microsoft’s “intelligence” layer built on top of the Microsoft Graph that mines customer data for Copilot. Starting June 16, the Work IQ API will require Copilot Credits and users will be charged directly when they build their own agents or apps calling the Work IQ APIs. They’ll also be billed if they use a third-party agent that uses Microsoft 365 data through the Work IQ APIs.
