GitHub Copilot will retire three models November 16 — map every pinned workflow now

GitHub will [retire three Copilot models on November 16, 2026](https://github.blog/changelog/2026‑10‑15‑upcoming‑deprecation‑of‑selected‑github‑copilot‑models‑in‑mid‑november/): Claude Opus 4.8, Claude Opus 5 and GPT‑5.6 Terra. GitHub says the change covers Copilot Chat, inline edits, ask and agent modes, and code completions. The suggested replacements are Claude Opus 5.5 for both Claude models, and either GPT‑6.1 Sol or GPT‑6 Luna for GPT‑5.6 Terra. That table is a routing recommendation, not evidence that the replacements behave identically. Model changes can alter tool use, latency, output length, code style and the amount of review a task needs. Treat November 16 as a production migration date wherever a model is explicitly selected. Inventory editor settings, repository instructions, custom agents, automation wrappers, training material and test fixtures. A selector that silently falls back is still a behavior change if nobody has checked the resulting code.

GitHub’s notice keeps Claude Opus 4.8 available to individual Copilot subscribers on annual plans so those customers retain an Opus offering. The exception does not extend to Claude Opus 5, GPT‑5.6 Terra, monthly individual subscriptions or organization plans unless GitHub documents a separate entitlement. That narrow exception makes account type part of the migration. Do not infer availability from a colleague’s model selector. Record the subscription owner, plan type and client where the model appears, then compare it with GitHub’s current [supported‑model table](https://docs.github.com/en/copilot/reference/ai‑models/supported‑models). Teams should plan around the common retirement date even if one developer can keep using Opus 4.8. Shared workflows need a model that every intended operator can select under the organization’s policy.

For Copilot Business and Enterprise customers using default model enablement, GitHub says the suggested alternatives will be enabled automatically. That convenience has two exceptions: an administrator may have disabled the global default, or may have explicitly disabled a replacement model. An automatic model‑list change therefore does not prove the team can run its replacement. Before changing instructions or test baselines, verify the model policy and confirm the model appears in each supported client. GitHub’s [model‑switching guide](https://docs.github.com/en/copilot/how‑tos/copilot‑in‑your‑ide/chat‑with‑copilot/change‑the‑chat‑model) also notes that availability depends on plan and surface, and that Copilot Extensions may override a user’s chat selection. Separate three checks: organization permission, client support and workflow configuration. Passing one does not establish the other two.

A manual chat can be moved with the model dropdown, but long‑lived dependencies often sit elsewhere. Search repositories and internal documentation for the retired model names and provider IDs. Review custom‑agent profiles, saved prompts, screenshots, onboarding steps, evaluation fixtures and any automation that validates a response against model‑specific assumptions. For each hit, classify the dependency. A cosmetic mention needs documentation cleanup. An explicit API or client selector needs a configuration change. A benchmark baseline or approval policy needs a controlled retest because the model affects the evidence behind the decision. Do not bulk‑replace names without checking the surrounding task. GitHub offers two alternatives for Terra because the appropriate choice depends on the workload. Sol may suit a demanding code task; Luna may be the better fit where latency and cost matter. Preserve the intended outcome and acceptance test rather than treating model names as interchangeable strings.

Choose representative work from the actual queue: a code explanation, a multi‑file edit, an agent task with tools and a completion‑sensitive workflow. Run the current and candidate model with the same context, then compare correctness, edits, tool calls, latency, review effort and credit consumption. Retain failures and unsupported claims rather than averaging them away. Update the default only after the candidate passes the checks that matter for that workflow. Keep a rollback path until November 16, and tell users which model to select when a client does not inherit the organization default. After the cutoff, verify that no scheduled job or onboarding document still promises the retired option. GitHub says removal itself requires no customer action; the platform will take the old models away. The useful work is earlier: finding every place that expects them and proving the replacement meets the same reader or developer job. Follow further platform changes on the [GitHub organization page](https://opentools.ai/organizations/github). *Image: Documentary capture of GitHub’s official October 15 deprecation notice, reproduced without generative alteration or artificial upscaling.*