Models, Standard

GitHub Copilot Deprecated Models: What Left on November 16, 2026, and What to Use Instead

Three cracked, fading model orbs flowing into one radiant shared platform — the deprecated GitHub Copilot models and their replacements.

On November 16, 2026, the GitHub Copilot deprecated models — Claude Opus 4.8, Claude Opus 5, and GPT-5.6 Terra — stopped being available across most Copilot experiences. GitHub announced the change on October 15, giving developers 32 days to switch, and published a one-to-one replacement for each model. If you use Copilot daily, the practical question is simple: which model should you pick now, and do you need to do anything at all?

The GitHub Copilot deprecated models: what left on November 16, 2026

Three models left GitHub Copilot on November 16, 2026. GitHub’s changelog entry lists each one with its suggested replacement:

Deprecated modelDeprecation dateSuggested alternative
Claude Opus 4.82026-11-16Claude Opus 5.5
Claude Opus 52026-11-16Claude Opus 5.5
GPT-5.6 Terra2026-11-16GPT-6.1 Sol or GPT-6 Luna

The removal covers most Copilot experiences, not just chat. GitHub names Copilot Chat, inline edits, ask and agent modes, and code completions. Two documented exceptions sit inside that scope: Claude Opus 4.8 stays available to individual GitHub Copilot subscribers on annual plans, and Claude Opus 5 remains available for runtime refusal fallback through existing authorized access. The deprecations were announced on October 15, 2026, which gave users 32 days between the announcement and the removal date.

For reference, the mapping in compact form:

{
  "deprecated": {
    "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"
  },
  "deprecation_date": "2026-11-16",
  "announced": "2026-10-15"
}

What to use instead of each deprecated model

Three dim channels flowing into bright open pathways, showing each deprecated Copilot model mapping to its replacement.

GitHub’s answer is straightforward. Both Claude models move to Claude Opus 5.5, and GPT-5.6 Terra moves to GPT-6.1 Sol or GPT-6 Luna. That is the mapping GitHub publishes; which replacement is actually available to you depends on your Copilot plan and where you are using Copilot, so the same suggestion does not land identically for every user.

The switch is not a price increase. GitHub’s rate card lists Claude Opus 5.5 at $4 input and $20 output per million tokens, against $5 input and $25 output for Claude Opus 4.8 and Claude Opus 5 — so the replacement costs less per token than either model it replaces, as reported from GitHub’s rate card.

What GitHub has not said is how the replacements compare in capability. GitHub’s announcement gives no comparison of Claude Opus 5.5 against Claude Opus 5, and no statement of how GPT-6.1 Sol and GPT-6 Luna differ from each other or from GPT-5.6 Terra. If your workflow depended on a specific strength of one of the deprecated models, the honest position is that GitHub has not documented whether the replacement matches it — you will find out by using it, not by reading the announcement.

What actually changes inside Copilot

The three models do not all disappear in the same way. Claude Opus 5 has been removed from the model picker, but GitHub says it remains available for runtime refusal fallback through existing authorized access, and existing enterprise model policies continue to apply. In practice, that means you cannot select Opus 5 any more, but systems that were already authorized to fall back to it keep that fallback.

Claude Opus 4.8 has the only exception in the notice. GitHub states that it is still available to individual GitHub Copilot subscribers on annual plans. That exception covers Opus 4.8 only — it does not cover Claude Opus 5 or GPT-5.6 Terra, and it does not cover Business or Enterprise seats. If you are on a monthly individual plan, a team plan, or a company plan, the annual-plan exception for Opus 4.8 does not apply to you — you lose the ability to select all three models, though the Opus 5 runtime refusal fallback through existing authorized access still stands.

One thing does not change: GitHub says no action is required to remove the deprecated models. You do not need to delete anything, uninstall anything, or clean up a configuration. The removal happens on GitHub’s side.

It is also worth being precise about what kind of change this is. This is a Copilot product change, not a vendor deprecation. Anthropic still lists Claude Opus 5 and Claude Opus 4.8 as active on its own API — with tentative retirement no sooner than July 24, 2027 for Opus 5 and no sooner than May 28, 2027 for Opus 4.8 — and OpenAI’s deprecations page does not list GPT-5.6 Terra as deprecated. All three models remain callable from the vendors’ own APIs after November 16; what has closed in Copilot is the ordinary route to select and use them, apart from the two documented exceptions — Opus 4.8 for annual individual subscribers and the Opus 5 runtime refusal fallback.

What individual Copilot users need to do

For most individual users, the work is a one-minute check. Open the model selector in Copilot Chat and confirm which model is selected. If it still names one of the three deprecated models, pick the suggested alternative: Claude Opus 5.5 in place of either Claude model, or GPT-6.1 Sol or GPT-6 Luna in place of GPT-5.6 Terra. Once enabled, the model appears in the Copilot Chat model selector in supported Copilot experiences.

If you are an individual subscriber on an annual plan, Claude Opus 4.8 remains available to you, so you keep an Opus option even without switching. GitHub gives no end date for that exception, so treat it as a reprieve rather than a permanent arrangement — the model is still listed as deprecated, and the exception could close in a later wave.

To confirm which models are available to your account, GitHub’s documentation on the models available in Copilot is the reference list. Availability depends on your Copilot plan and where you are using Copilot, such as GitHub.com or an IDE, and GitHub states that availability is subject to change.

If you still need Claude Opus 5 or GPT-5.6 Terra inside Copilot itself, one reported route is to add the model with your own key — for example as a custom endpoint model in VS Code — since the models remain available on their vendors’ APIs. That is a workaround, not something GitHub recommends in the deprecation notice.

What Copilot Enterprise and Business admins need to do

For Copilot Enterprise and Copilot Business customers, the suggested alternatives are automatically enabled under default model enablement — unless an administrator has turned off the global default or explicitly disabled the model. In that case, the alternatives do not appear until an admin enables them through the models’ model policies in Copilot settings.

That default behaviour is part of a wider policy. For enterprises on Copilot Business or Enterprise plans, a policy controls whether unconfigured generally available models are enabled or disabled by default, and a new model inherits the default until it is explicitly configured. Some models sit outside that policy entirely: pre-GA models, open-weight models, and models not covered by GitHub’s data retention agreement are disabled by default regardless of the default-availability setting.

The practical check for an admin is therefore twofold: confirm the global default is on, or confirm the replacement models are explicitly enabled in model policies — and confirm that the deprecated models are not still named in any workflow, integration, or model policy that assumes they exist. Existing enterprise model policies continue to apply; check each replacement model’s effective policy against the documented global default and the explicit model settings rather than assuming the old settings carry over. GitHub Enterprise customers with questions are encouraged to reach out to their account manager.

How to check which model you are on and switch

A glowing circular model selector dial with one slot lit up, showing how to check which Copilot model is selected.

The model selector lives in Copilot Chat in supported Copilot experiences. Open Chat, look at the model picker, and check the name shown. If it names Claude Opus 4.8, Claude Opus 5, or GPT-5.6 Terra, select the suggested replacement from the same picker. The replacement appears there once it is enabled for your account — under default enablement it is automatic for Enterprise and Business customers, and for individual users it depends on your plan and where you are using Copilot, so check the supported-models reference if it does not appear.

Two details from GitHub’s documentation are worth knowing while you check. First, the models available to you depend on your plan and on where you are using Copilot, so a model that appears in VS Code may not appear everywhere. Second, GitHub lists GPT-5.3-Codex as the long-term support model, used when no other models are available — a backstop that exists independently of this deprecation. If a replacement model does not appear in your picker, check your plan, the client you are using, and whether the model is enabled for your account before assuming it is disabled.

What GitHub has not said yet

Four gaps in the announcement are worth stating plainly, because filling them in would be guessing.

First, GitHub gives no reason for each deprecation beyond the lifecycle change itself. There is no statement that the models were underperforming, too expensive, or superseded on capability — only that they are deprecated and here are the replacements.

Second, the Claude Opus 4.8 annual-plan exception has no end date. GitHub does not say when, or whether, it will close.

Third, GitHub does not explain how GPT-6.1 Sol and GPT-6 Luna differ from each other or from GPT-5.6 Terra. Both are offered as the replacement for Terra with no guidance on which to choose, and the announcement itself carries no comparison.

Fourth, GitHub does not specify what happens to a request that names a retired model after November 16 — whether it fails, falls back, or is silently remapped.

One pattern is visible, though. This is the third retirement wave in about six weeks: a wave on October 2 removed Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7, and a wave on October 19 removed six more models including GPT-5.5, GPT-5.4 and Grok 4.5. Three dated waves do not prove a recurring cadence, but they do show the list moves often, and GitHub’s own supported-models page now carries a table of models retired or scheduled for retirement. The durable habit is not to memorize the current list but to check that page — and your model picker — before assuming the model you used last month is still there.