deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

GitHub Copilot to remove Claude Opus 4.8, Opus 5 and GPT-5.6 Terra on November 16

GitHub Copilot will remove Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra from chat, completions, inline edits and agent modes on November 16, 2026, with named replacement models suggested.

GitHub Copilot to remove Claude Opus 4.8, Opus 5 and GPT-5.6 Terra on November 16

Three models, every surface

On November 16, 2026, GitHub Copilot will drop Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra from all of its experiences. According to a post on dev.to that works through a changelog entry GitHub published on October 15, the removal covers not just the chat model picker but also inline edits, ask and agent modes, and code completions. The changelog asks users to move their workflows and integrations to supported models before the deadline.

The replacement mapping

The changelog table reproduced on dev.to lists Claude Opus 5.5 as the suggested replacement for Claude Opus 4.8, and GPT-6.1 Sol or GPT-6 Luna for both Claude Opus 5 and GPT-5.6 Terra. The dev.to post's own summary, however, points both Opus models at Opus 5.5 — a discrepancy worth resolving against the changelog itself before planning a migration.

There is one carve-out: Claude Opus 4.8 remains available to individual Copilot subscribers on annual plans, which GitHub frames as preserving an Opus option for that group. Monthly individual subscribers and seats managed through Copilot Business or Copilot Enterprise are not covered, and no plan keeps Opus 5 or GPT-5.6 Terra. The dev.to author's recommendation even for those who qualify is to migrate anyway, rather than split a team across two model generations.

Where the breakage hides

Chat pickers are the easy part to fix. The dev.to post argues the real exposure sits in unattended configuration: prompt files, CI workflows, Copilot CLI scripts, custom agent definitions, workspace settings that store a default model, and internal runbooks that tell people which model to pick. It recommends sweeping repositories with grep patterns covering the old model names and treating every hit as a defect until it points at one of the replacements.

GitHub has not documented what happens when a request names a retired model after the cutoff. The dev.to post declines to guess between outright failure and silent rerouting to a default, and suggests the safe engineering assumption is that any pinned reference simply breaks.

Policy checks and testing

For Copilot Business and Enterprise customers, the replacement models are enabled automatically under default model enablement, unless an administrator has turned off the global default or disabled a specific model. The dev.to post stresses confirming in org model policies that a non-admin user can actually select Opus 5.5, GPT-6.1 Sol and GPT-6 Luna, because a model nobody can select does not count as a replacement.

Behaviour is the other open question. Prompts tuned against one model will not necessarily perform the same on its successor, and the window before November 16 is the chance to run old and new models side by side on a team's most-used prompts and compare the results. For Terra users facing a choice between Sol and Luna, the post advises comparing output quality on real tasks rather than defaulting to whichever is cheaper. The removal itself needs no work: GitHub says it retires the models on its own once they are deprecated.

Why it matters

The dev.to author notes this is the third wave of Copilot model retirements in 2026 on a similar cadence, and draws a broader lesson: any model name hard-coded into a workflow has only months of useful life, with the changelog serving as the sole advance notice. The immediate consequence is operational — an agent definition, CI script or IDE setting pinned to one of the three retiring names should be presumed broken after November 16 unless it is found and updated first. The longer-term consequence is a maintenance practice: treating model identifiers as perishable configuration to be audited regularly, not permanent infrastructure to be set once and forgotten.

  • #github-copilot
  • #developer-tools
  • #ai-models
  • #deprecation
  • #llm

Related posts