GitHub Adds xAI's Grok 4.5 Reasoning Model to Copilot

GitHub has started rolling out xAI's Grok 4.5 as a selectable model inside GitHub Copilot, adding another high-context reasoning option to the coding assistant's model picker. The July 28 changelog says the model is aimed at fast agentic coding and complex multi-step software work, with support for text and image inputs and a context window of up to 500,000 tokens.

The update matters because Copilot is increasingly a routing layer for competing AI models rather than a single assistant tied to one model family. Developers and administrators can now choose between different models for different jobs, weighing latency, reasoning behavior, context size and price. GitHub says Grok 4.5 performed strongly in its internal testing on terminal-based coding tasks in Visual Studio Code and Copilot CLI, especially when using tools in parallel and taking direct action during exploratory work.

Where it is available

According to GitHub, Grok 4.5 is rolling out gradually to Copilot Pro, Pro+, Max, Business and Enterprise plans. The model can be selected in Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot cloud agent, the Copilot app, JetBrains IDEs, Xcode and Eclipse. Users who do not see it immediately may need to wait for the staged rollout.

For organizations, the release adds an administrative control rather than silently enabling a new third-party model across a company. GitHub says Copilot Business and Enterprise administrators must turn on the Grok 4.5 policy in Copilot settings before users can access it, and that the policy is off by default.

Why developers should notice

• The model gives Copilot users another option for long-context coding sessions and agent-style workflows.
• Business and Enterprise access depends on an explicit administrator policy decision.
• Billing follows provider list pricing under usage-based billing, so teams will need to watch costs as model choice expands.

The announcement is not a benchmark paper and GitHub frames its performance comments as internal testing, so the practical impact will depend on how the model behaves in real repositories. Still, the addition shows how quickly coding assistants are becoming multi-model workbenches, where developers choose specialized models for planning, debugging, command-line work or large-codebase exploration.

Sources

• github.blog

Cover photo by Daniil Komov on Pexels, used under the Pexels License.