GitHub Copilot removes three major AI models on November 16
Copilot is dropping Claude Opus 5 and GPT-5.6 Terra from its menu. The models are fine, but your workflows will break if you rely on them.
Tools · Source: GitHub Changelog
What happened
GitHub is cutting three major AI models from Copilot on November 16, 2026. Claude Opus 4.8, Claude Opus 5, and GPT-5.6 Terra are leaving the menu. This deprecation impacts every part of the Copilot experience. It hits Copilot Chat, inline edits, agent mode, and code completions. If you use VS Code, JetBrains IDEs, or the Copilot Workspace web interface, your model list is about to shrink.
GitHub suggests moving to newer models immediately. The official alternatives are Claude Opus 5.5, GPT-6.1 Sol, or GPT-6 Luna. Enterprise and Business customers will see these alternatives enabled automatically under default settings. Administrators who explicitly disabled global defaults must manually update their model policies in Copilot settings. Users will then see the new models in their chat selector.
There is one narrow exception to this purge. Individual developers on annual Copilot subscriptions get to keep Claude Opus 4.8. Everyone else loses access to all three models on the cutoff date. This includes monthly individual users, team seats, and enterprise licenses. GitHub is clearing capacity and licensing costs on its larger user base while avoiding broken workflows for long-term individual subscribers.
Key facts
- November 16, 2026 — Date GitHub removes the three models from Copilot
- Claude Opus 5.5 — Suggested replacement for Opus 4.8 and Opus 5
- 816 billion — Total parameters in Meta new Muse Spark Open model
- 50 — Number of linked repositories Cursor 0.45 can map
Why it matters
The models themselves are not dead. They are just losing their route through GitHub Copilot. If you built custom prompts, extensions, or evaluation pipelines tuned specifically to Opus 5 or Terra 5.6, your setup is in trouble. Any CI job that explicitly calls one of those three model names by ID will fail or silently fall back to whatever default GitHub assigns. You have to bet on the model, not the menu. Running a long list of models is expensive for GitHub due to inference costs and support overhead. They are trimming the fat.
This contraction happens exactly as the coding agent market explodes with new options. Cursor just shipped version 0.45 with cross-repository dependency mapping for up to 50 linked repositories. OpenAI updated the Codex CLI three times in two weeks, making GPT-6.1 Sol the default and adding instant steering. Meanwhile, Meta dropped Muse Spark Open, a massive 816-billion parameter open-weight model optimized for coding. GitHub is pushing users toward task-based automation, like the new natural-language debugging mode in Copilot Workspace, rather than manual model selection. But competitors are giving developers more control, not less.
For builders
Audit your CI jobs and extensions
Any workflow or extension explicitly calling the retired model IDs will fail or fall back to a default on November 16. Engineering teams must update their integrations to use Claude Opus 5.5 or GPT-6.1 Sol before the cutoff. Teams paying for enterprise seats lose access entirely, so the engineering department pays the price in broken pipelines if they ignore this deadline.
Evaluate alternative coding agents
Copilot is shrinking its lineup while competitors expand their capabilities. Teams should test Cursor 0.45 or OpenAI Codex CLI for complex polyglot codebases. If GitHub no longer supports your preferred model, you might find better routing and fewer restrictions by moving your developer seats to a rival platform.
Leverage open weight coding models
The open-weight ecosystem is moving incredibly fast right now. Meta just released Muse Spark Open with a one-million-token context window, and Reflection AI launched Beam. Founders building internal coding tools can bypass platform restrictions entirely by hosting these models on their own infrastructure.
My take
I always say you should bet on the model, not the menu. Never trust a platform to keep your favorite model alive forever. GitHub is cutting their overhead while competitors are shipping massive updates. Build your workflows so you can swap out the routing layer the second a provider pulls the plug.
Original reporting: GitHub Changelog. This is my rewrite and opinion.