OutYet reporting
Microsoft 365 makes GPT-5.6 Sol a workflow decision, not just an API choice
OpenAI says GPT-5.6 will become Microsoft 365 Copilot's preferred model, extending the practical importance of its Sol, Terra, and Luna tiers beyond direct API evaluation.
OpenAI said on July 9 that GPT-5.6 will become the preferred model in Microsoft 365 Copilot, covering Word, Excel, PowerPoint, Chat, and Cowork. Its announcement says Microsoft will access the models through the OpenAI API in addition to serving them natively. This is a product-integration update, not independent evidence of model quality, but it places the GPT-5.6 family in a consequential enterprise surface where drafting, analysis, presentations, and collaboration are already part of the user workflow.
The timing matters because OpenAI's broader GPT-5.6 material presents a three-tier family rather than a single flagship: Sol is the flagship, Terra is positioned as a lower-cost model, and Luna as the fastest and least expensive tier. The Microsoft 365 announcement discusses GPT-5.6 at the family level, while the larger product page describes different access and effort options across ChatGPT, Codex, and the API. For technical teams, that makes the integration a reason to distinguish product-level model selection from a single benchmark result.
OpenAI compares Sol with GPT-5.5 and named competitors in its own evaluation tables, and says Sol reaches 80 on the Artificial Analysis Coding Agent Index with its max reasoning setting. It also says GPT-5.6 can use Programmatic Tool Calling in the Responses API and that its ultra setting coordinates four agents in parallel by default. Those are useful implementation details, but they are vendor-reported performance and product claims: they depend on the specified reasoning setting, task, tool configuration, and cost assumptions, so they should not be treated as a universal ranking.
For model users, the practical question is whether a Copilot workflow preserves the controls needed to validate outputs, manage data access, and choose an appropriate tier for the task. OpenAI's material describes direct API access, tiered pricing, and tool-oriented features, but the cited announcements do not independently establish accuracy, latency, regional availability, tenant configuration, or the effect of the change in a particular Microsoft 365 deployment. Teams should therefore test representative document, spreadsheet, and presentation workloads rather than infer production suitability from the integration announcement alone.