GitHub Copilot retires Opus 4.8, Opus 5 and GPT-5.6 Terra
Edited and reviewed by AI Ping Editorial Team
3-line summary
- GitHub Copilot retired Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra across most features on November 16.
- Individual annual subscribers retain access to Opus 4.8. Work accounts see replacements according to their organization’s model policies.
- GitHub lists Opus 5.5, GPT-6.1 Sol and GPT-6 Luna as alternatives. Chat and code completions have separate model settings.
GitHub retired Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra across most GitHub Copilot features on November 16. It said the change covers Copilot Chat, inline edits, ask and agent modes, and code completions. Individual annual subscribers can keep using Opus 4.8.
GitHub first listed the retirement date in its October 15 notice and confirmed that it had carried out the change on November 16. It points to Claude Opus 5.5 as an alternative for both retired Opus versions. For GPT-5.6 Terra, it names GPT-6.1 Sol or GPT-6 Luna.
Exceptions for annual plans and a runtime fallback
In its October notice, GitHub said it would retain Opus 4.8 so individual annual subscribers would still have an Opus option in Copilot. The November retirement notice confirms that exception. Annual Pro and Pro+ subscribers still on request-based billing use model-specific multipliers for those requests.
GitHub removed Opus 5 from the model picker. The company says Copilot retains it for a limited fallback if another model refuses a request, subject to existing access permissions and enterprise model policies. GitHub says it handles the removal of retired models from the picker.
Which plans offer the replacements
GitHub added Opus 5.5 and GPT-6 Luna to Copilot on September 22 and GPT-6.1 Sol on September 29. Opus 5.5 and GPT-6.1 Sol are available on Pro+, Max, Business and Enterprise. Luna is also available on Pro. A separate model, GPT-6.1 Terra, joined Copilot on October 26. The version covered by the November retirement is GPT-5.6 Terra.
On plans that use AI credits, the model and the number of tokens it processes affect how quickly a chat draws down the allowance. GitHub lists standard, uncached rates per million input and output tokens of $0.10 and $0.50 for Luna, $2 and $10 for Sol, and $4 and $20 for Opus 5.5. Sol and Luna have a separate rate for inputs over 272,000 tokens. Paid plans still offer inline code completions and next edit suggestions without consuming AI credits.
Each AI credit is worth one cent. Individual plans come with different allowances. In Business and Enterprise, GitHub pools the credits allocated to users under the same billing account, so the organization’s total usage and its mix of models affect how quickly it uses the included credits.
GitHub says Business and Enterprise accounts get the replacement models automatically when the default model policy allows them. The administration guide says that default covers generally available models without their own policy. An enterprise administrator can block a specific model or leave access decisions to individual organizations. Colleagues on the same plan can therefore see different models in Chat based on their organization’s settings.
Chat and inline suggestions have separate settings
GitHub’s VS Code documentation says people can switch models in Copilot Chat while a conversation is in progress. The model they select generates replies to subsequent prompts. The editor uses a separate setting for inline suggestions that continue a line of code. Changing the model in Chat leaves that completion selection in place.
GitHub’s instructions for completions say the latest VS Code and Copilot extension can show multiple options when an alternative is available. If only one model is available for inline suggestions, the picker shows just that one. Choices in Chat also vary by subscription and client. GitHub adds that Copilot Business members must have permission from their organization to change the Chat model.
Custom agents can have a separate model in their saved profile. GitHub’s reference says the named model is used when that agent runs. A profile without a model entry inherits the default. Changing the model for a chat in the same repository leaves a separately named model in a profile untouched. GitHub’s retirement notice also asks users to update workflows and integrations that refer to retired models.
Sources
- November 16 retirement confirmation — GitHub
- Claude Opus 5.5 availability — GitHub
- GPT-6.1 Sol availability — GitHub
- GPT-6 Luna availability — GitHub
- GPT-6.1 Terra availability — GitHub
- Copilot models, token prices and annual billing — GitHub Docs
- Enterprise model availability policies — GitHub Docs
- Change the Copilot Chat model — GitHub Docs
- Change the inline suggestion model — GitHub Docs
- Custom agent profile model selection — GitHub Docs