On July 28, 2026, GitHub announced that xAI's Grok 4.5 is beginning to roll out in GitHub Copilot. GitHub describes it as a reasoning model for fast agentic coding and complex multi-step workflows, with a context window of up to 500,000 tokens, text and image input, and low, medium, and high reasoning effort.
Availability is a major part of the update. Grok 4.5 will appear gradually in the model picker across Visual Studio Code, Visual Studio, Copilot CLI, Copilot cloud agent, the Copilot app, JetBrains, Xcode, and Eclipse. It is listed for Copilot Pro, Pro+, Max, Business, and Enterprise. GitHub says the rollout is gradual, so accounts will not necessarily see the model at the same time.
GitHub also says its internal testing showed strong results on terminal-based coding tasks in Visual Studio Code and Copilot CLI, especially when the model dispatched tools in parallel and took direct action. That is GitHub's description of internal testing, not a public, reproducible independent benchmark. Results will still depend on repository structure, tool permissions, prompts, and verification steps.
For Business and Enterprise administrators, availability is not the same as open access. GitHub says administrators must enable the Grok 4.5 policy in Copilot settings, and that policy is off by default. This puts model choice inside enterprise governance, giving teams a chance to consider data, compliance, cost, and agent behavior before adding it to the model picker.
Cost is another boundary. GitHub says Grok 4.5 is billed at provider list pricing under usage-based billing. A 500,000-token context limit may reduce repeated background transfer in long or parallel agent sessions, but it does not guarantee lower cost or better accuracy for every task. Teams should measure actual tokens, tool calls, retries, and human review time.
The update shows how model competition is reaching users through a governed coding platform rather than only through standalone chat. The useful evaluation question is not whether a new name appears in the model picker. It is whether the model can work within constrained permissions, leave inspectable work records, and stop at a human approval point when it fails or reaches a risky action.
The rollout, enterprise policy, and usage-based billing all make “available” a conditional statement. Development teams considering adoption should test code quality, tool use, data boundaries, cost, and rollback procedures, then place GitHub's internal testing claims inside a broader evidence chain.
