Microsoft sets GPT-5.6 default

Microsoft has made OpenAI’s GPT-5.6 Sol the default model for its own staff using GitHub Copilot, according to a CNBC report published on August 5. The report said Jay Parikh, Microsoft’s executive vice president for CoreAI, told employees in an internal memo to use the model as part of an efficiency push. (cnbc.com) CNBC reported that Microsoft framed the move around rights it holds through its partnership with OpenAI. The change applies to Microsoft’s internal staff deployments, not to all GitHub Copilot customers. (github.blog) ### Is this a GitHub Copilot change for everyone? GitHub said on July 9 that OpenAI’s GPT-5.6 Sol, Terra and Luna were available in GitHub Copilot, but only if Copilot Enterprise and Copilot Business administrators enabled the policy. (cnbc.com) GitHub’s changelog said that policy was off by default for those customer plans. That means Microsoft’s internal decision is separate from the product’s default settings for outside companies. CNBC reported that Microsoft’s instruction was aimed at its own employees. Windows Forum, citing the same CNBC reporting, said the distinction matters because the internal policy does not automatically change what outside IT teams see in their Copilot environments. (cnbc.com) ### Why would Microsoft steer staff to one model? CNBC reported that Microsoft is leaning on intellectual-property rights from its OpenAI partnership and pushing employees toward the company’s top model as part of an efficiency effort. The report described the directive as part of Microsoft’s attempt to get more value from AI use inside the company. (github.blog) Jay Parikh also tied internal AI use to cost discipline, according to separate reporting on internal guidance. PCMag and other outlets, citing reports on Microsoft’s memo, said Parikh told employees that “tokenmaxxing” was not the goal and that teams should focus on impact from AI use rather than raw consumption. (cnbc.com) ### What does “tokenmaxxing” tell us about the rollout? Microsoft has begun treating AI usage as a managed internal resource, according to reports published this week. Yahoo Finance, republishing TechRadar reporting, said Microsoft’s internal guidance referred to an “AI token budget target” for divisions as of July 2026 and said some engineers were spending from hundreds to a few thousand dollars a month in tokens. (cnbc.com) The Register, citing the same internal email, reported that divisions could face restrictions if usage ran too high. That suggests the model change is arriving alongside spending controls, not as a blanket invitation to use more AI without limits. (pcmag.com) ### How does this fit with Microsoft’s broader Copilot strategy? Microsoft has already been moving GPT-5.6 into other products. On July 9, Microsoft said GPT-5.6 became available in Microsoft 365 Copilot and called it the preferred model for that service. Microsoft said it would both serve the models natively and access OpenAI models directly through the API for Microsoft 365 customers. (finance.yahoo.com) OpenAI’s GPT-5.6 family also reached GitHub Copilot in July. GitHub said the rollout would be gradual across Visual Studio Code, Visual Studio, Copilot CLI, GitHub’s cloud agent, GitHub.com, mobile apps and several IDEs. (theregister.com) ### What should readers watch next? GitHub’s July 9 changelog said administrators on Copilot Enterprise and Copilot Business still need to enable GPT-5.6 access in settings. That makes future admin guidance, usage policies and rollout notes the clearest public markers for whether Microsoft’s internal default spreads into broader enterprise practice. (techcommunity.microsoft.com) Microsoft’s internal guidance is also likely to stay in focus as managers enforce token budgets and usage targets. CNBC identified Parikh as the executive behind the memo, and additional changes would most likely show up through GitHub Copilot documentation, Microsoft product blogs or further internal-policy reporting. (cnbc.com) (github.blog)