GitHub retired Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra from most Copilot experiences on November 16. The change affects chat, inline edits, ask and agent modes, and code completions. It puts a previously announced retirement into effect, rather than introducing a new deadline for developers.
Replacement models and two exceptions
GitHub recommends Claude Opus 5.5 for users moving from either retired Opus model. Its suggested alternatives to GPT-5.6 Terra are GPT-6.1 Sol or GPT-6 Luna.
There are important limits to the retirement. Individual subscribers on annual plans retain access to Opus 4.8. Opus 5 disappears from the model picker but remains available for runtime refusal fallback through existing authorized access, with enterprise policies still applying. That fallback exception does not restore it as a manually selectable model.
Developers comparing options can consult the GitHub Copilot profile7.3 out of 10, but should check the live model list for their particular plan and client. GitHub’s documentation separates availability by product surface; a model listed for one experience is not a promise that every integration offers it.
Administrators should check model policies
For Business and Enterprise customers using default model enablement, GitHub says the suggested replacements become available automatically unless an administrator has disabled the default or the individual model. Organizations that opted out of the default need to enable replacements in their model policies.
GitHub’s access guide identifies three separate constraints: the subscription plan, the client being used and organizational restrictions. A developer who cannot see a replacement should therefore check those settings before assuming the retirement removed access to Copilot itself.
Cloud-agent sessions have their own selection rules
The cloud-agent guide allows model selection when starting tasks through supported entry points, including issue assignment on GitHub.com, the agents interface and GitHub Mobile. Where there is no picker, Copilot uses Auto. Models that support configurable reasoning expose an additional control; more reasoning can take longer and consume more AI credits.
For teams maintaining repeatable coding-agent workflows, the practical work is to check saved model choices, confirm the replacement is permitted, and rerun representative tasks. Retirement changes which model executes the request; it does not establish that a replacement will produce identical edits, latency or costs.