November 16, 2026 is a practical QA deadline for teams using GitHub Copilot. GitHub says it will retire Claude Opus 4.8, Claude Opus 5, and GPT-5.6 Terra across Copilot experiences, including Chat, inline edits, ask and agent modes, and code completions. GitHub recommends Claude Opus 5.5 for the two Claude models and GPT-6.1 Sol or GPT-6 Luna for GPT-5.6 Terra. A model picker update is not proof that an automation workflow still behaves as expected: treat this as a focused GitHub Copilot model changeover QA exercise.
What changes on November 16
| Retiring model | GitHub-suggested alternative |
|---|---|
| Claude Opus 4.8 | Claude Opus 5.5 |
| Claude Opus 5 | Claude Opus 5.5 |
| GPT-5.6 Terra | GPT-6.1 Sol or GPT-6 Luna |
GitHub announced this retirement on October 15, 2026. It says Claude Opus 4.8 remains available to individual Copilot subscribers on annual plans, while the broader retirement applies across Copilot. For Business and Enterprise, GitHub says suggested alternatives are automatically enabled only when default model enablement is still active; administrators who disabled that default or a specific model must check their model policies.
A compact QA readiness pack
- Inventory model-dependent work. Record every saved prompt, custom instruction, agent task, CI integration, and team playbook that specifies one of the retiring models.
- Freeze representative fixtures. Select a small, versioned set: a bug reproduction, a failing test repair, a test-generation request, a code-review pull request, and a repository-navigation task.
- Capture the old baseline now. Preserve prompt, repository commit, tool permissions, generated diff, test commands, runtime, failures, and human verdict. Do not compare only the assistant’s prose.
- Run candidates separately. Execute the same fixture with the suggested replacement model, from a clean checkout and the same permitted tools. Keep results labeled by model and date.
- Judge behavior independently. Verify the original failure, inspect the diff, run targeted negative checks and the normal CI suite. A passing generated test alone is not sufficient evidence.
- Check access before the deadline. Confirm the intended alternative appears in the supported Copilot clients and is permitted by enterprise, organization, and repository policy.
Why this matters for QA engineers
A replacement can be a good fit without producing the same approach, edits, latency, or tool calls as its predecessor. That variability matters when a team has prompts tuned around a specific model, relies on an agent to make test changes, or measures review outcomes across a release window. Preserve the baseline and evaluate outcomes such as correctness, unnecessary edits, test coverage, reproducibility, and CI result—not just acceptance rate or response quality.
Minimal comparison record
Fixture: checkout-validation-bug
Repository commit: 4f2d8c1
Retiring model: GPT-5.6 Terra
Candidate model: GPT-6.1 Sol
Original failure reproduced: yes/no
Generated files changed: list
Targeted tests: pass/fail
Full CI: URL + pass/fail
Independent reviewer verdict: accept/rework
Do not turn this into a benchmark claim
GitHub’s suggested alternatives are migration guidance, not a promise of identical outputs or universal quality gains. Keep scope modest: validate the workflows your QA group actually depends on, retain existing release gates, and document exceptions. If policy changes are needed, make them before the retirement date rather than discovering a missing model in the middle of an incident or release.
Bottom line
The November 16 Copilot retirement date is a useful forcing function for an evidence-led GitHub Copilot model changeover QA check. Snapshot a handful of real fixtures, test the approved alternative from a clean environment, and use independent CI and review evidence to decide whether the workflow is ready.