GitHub Copilot model picker showing Grok 4.5 as a selectable model Image: GitHub Blog / github.blog
by VibecodedThis

Grok 4.5 Is Now in GitHub Copilot's Model Picker

xAI's Grok 4.5 joined the Copilot roster on July 28, giving subscribers access to a 500K-context reasoning model built for fast agentic work.

Share

GitHub added Grok 4.5 to Copilot’s model picker on July 28, rolling the model out gradually across all eligible plans. The addition puts xAI’s coding-focused reasoning model alongside Claude Opus 5, OpenAI’s GPT-5.6 family, and Gemini in what has become one of the widest model selections of any AI coding tool.

Grok 4.5 is available to Copilot Pro, Pro+, Max, Business, and Enterprise subscribers. Business and Enterprise admins have to enable it manually — the model is off by default for those tiers, requiring a policy change in Copilot settings before users can select it.

What Grok 4.5 Brings

The model was built with agentic and terminal-based coding tasks in mind. Key specs:

  • Context window: 500,000 tokens
  • Inputs: text and images
  • Reasoning effort: low, medium, and high — selectable per request
  • Billing: usage-based at provider list pricing ($2 per million input tokens, $6 per million output)

GitHub says it performed well in internal testing on parallel tool dispatch and direct action execution, which makes it a reasonable fit for Copilot’s cloud agent and CLI workflows. It’s accessible from the model picker in VS Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, and GitHub’s web-based cloud agent.

Why This Matters

GitHub Copilot’s model picker has quietly become a meaningful battleground. A year ago, Copilot was essentially synonymous with a specific OpenAI model. Now it’s a routing layer across models from at least four different providers, and picking the right one for a task is a real consideration.

Grok 4.5 launched on July 8 at a price point well below Claude Opus 5 — $6 per million output tokens versus $75 — and xAI’s own benchmarks showed competitive performance on several coding tasks, though results are mixed across benchmarks. The token efficiency gap is notable: xAI reported Grok 4.5 resolving tasks at roughly a quarter of the output token count that Opus 4.8 used on the same SWE-Bench Pro runs. Whether that holds across your actual codebase is a different question, but it’s worth testing.

For Copilot users who want to try it, the model picker is the starting point. For Business and Enterprise, check with your admin first.

Sources:

Share