grok 4.5 lands in copilot, but enterprise admins hold the switch

GitHub started rolling Grok 4.5 into GitHub Copilot on July 28, 2026. the model is pitched for fast agentic coding and multi-step workflows, with a context window up to 500,000 tokens, text and image input, and low, medium, or high reasoning effort settings.

availability is broad but not automatic everywhere. Copilot Pro, Pro+, Max, Business, and Enterprise users are in scope, and GitHub lists VS Code, Visual Studio, Copilot CLI, Copilot cloud agent, the Copilot app, JetBrains, Xcode, and Eclipse as target surfaces. the rollout is gradual, and Business/Enterprise admins have to enable the Grok 4.5 policy before their users can select it.

the important part is the admin switch

the model-picker story is no longer just "which model is smartest this week." Copilot is becoming a governed model marketplace inside the dev workflow: enterprise admins decide which provider models are allowed, usage-based billing applies provider list pricing, and the same editor can expose Microsoft, Anthropic, Google, xAI, and open-weight options.

that is good for teams that want one control plane. it is also a reminder that third-party models inside Copilot are not neutral toggles. each one brings cost, data-policy, reliability, and procurement questions. GitHub making Grok 4.5 off by default for Business and Enterprise is the responsible shape.

github also tightened the measurement loop

the same day, GitHub expanded Copilot app usage metrics across its reporting rollups. Copilot app activity can now be attributed to individual users in enterprise-user and organization-user reports, and app activity can be broken out by feature, model, and language next to IDE, chat, code review, and coding-agent surfaces.

that matters because the Copilot app and cloud-agent loop are moving work out of the editor sidebar and into delegated sessions. without user-level and model-level metrics, leaders only know that "someone used the app." with the new fields, they can ask which teams are adopting it, which models they are burning through, and whether app-generated code is showing up in accepted code activity.

honest read: Grok 4.5 gives Copilot another serious frontier-ish model in the picker, but the durable news is GitHub pairing model sprawl with governance and telemetry. the winners in enterprise coding agents will not just have the best model button. they will have the boring controls that make finance, security, and engineering managers say yes.