The coding-agent race is moving from benchmark pages into the interfaces where developers already review, change, and ship code. GitHub said on July 24, 2026 that Claude Opus 5 is now available in GitHub Copilot, giving eligible users a new model choice for complex coding work inside the Copilot model picker.

The distinction matters because Anthropic’s model release and GitHub’s distribution surface are different events. The model itself was already news; GitHub’s changelog makes it available where many teams actually assign agentic coding tasks. GitHub describes the target workload as complex, long-running coding that needs careful reasoning, effective tool use, and reliable execution over multiple steps.

In GitHub’s early testing, Opus 5 showed strength on autonomous code changes, regression verification, and tasks that coordinate multiple tools. That puts the model in the part of Copilot that is less about autocomplete and more about the full loop: identify the change, edit the code, run checks, interpret failures, and keep the work bounded enough for review.

The initial availability list is broad. Claude Opus 5 will be available to Copilot Pro+, Max, Business, and Enterprise users, with selection in Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot cloud agent, the Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains, Xcode, and Eclipse. GitHub says the rollout is gradual, so model visibility may lag account eligibility.

For organizations, the important control is policy. Copilot Enterprise and Copilot Business administrators must enable the Claude Opus 5 policy in Copilot settings before users can access it. That keeps the rollout from becoming an unmanaged model switch across the company. Billing also follows usage-based pricing at the provider API list price, which will push teams to reserve the model for work where stronger reasoning offsets the cost.

Security behavior is another practical detail. GitHub notes that Opus 5 includes enhanced safeguards for high-harm cyber content, and that some cyber or security-adjacent requests may be blocked. Developers working on legitimate security tasks may need to add benign context or select a different available model when the request trips those controls.

The bigger signal is that Copilot is becoming a model-routing layer for software work. GitHub can now put different models behind different task profiles instead of presenting Copilot as one assistant. The next useful evidence will be operational: how often Opus 5 reduces failed agent runs, whether regression verification improves, and how teams balance model cost against fewer review cycles.

#claude-opus-5

Claude Opus 5 enters GitHub Copilot for long-running coding agents