Claude Opus 4.5 is in public preview for GitHub Copilot (github.blog)

Anthropic’s Claude Opus 4.5 is now in public preview inside GitHub Copilot, available across Copilot Chat in VS Code (Agent, Plan, Ask, Edit), github.com, GitHub Mobile (iOS/Android) and the Copilot CLI. Early internal tests report that Opus 4.5 outperforms previous coding benchmarks while cutting token usage roughly in half, and it will be the default Copilot coding agent during the promotional period. Rollout is gradual; admins on Enterprise & Business plans must enable the Claude Opus 4.5 policy, while Pro/Pro+ users can pick the model in VS Code after confirming a one-time prompt. For the AI/ML and developer community this matters because the combination of higher coding accuracy and substantially lower token consumption directly improves responsiveness, cost efficiency and the practicality of longer or more interactive coding sessions. Halving token usage reduces compute and billing overhead for hosted Copilot interactions and can enable larger context windows or more frequent assistant turns without raising costs. Broad integration across web, editor, mobile and CLI also makes it easy to benchmark Opus 4.5 in real-world workflows and supply targeted feedback to GitHub’s community as the preview expands.